It’s a beautiful time to be in content marketing. AI writes your blog posts. It drafts your meta descriptions. It can knock out a month of social captions before your coffee goes cold.
There’s just one problem: it all sounds the same. Yours, theirs, and the competitor you don’t even respect.
Same tidy sentences, same agreeable rhythm, same faint smell of nobody in particular. You can publish a hundred pages a month and still sound like a brand in witness protection.
I taught an SMX Master Class this spring on scaling content with Claude. The question that kept coming up had nothing to do with keywords or rank tracking: How do we get it to sound like us?
The answer is a Claude brand skill: a structured set of voice, tone, visual, and formatting rules that teach Claude how your brand should sound before it writes a single word.
Here’s how to build one that keeps your AI-assisted content recognizable, consistent, and distinctly yours.
What a Claude brand skill actually does
A Claude brand skill is your brand’s walkout song. When I played professional soccer, mine was “O Fortuna” from Carmina Burana because apparently I wanted to enter the field like a Roman army with shin guards.
Ridiculous? Maybe. Memorable? Absolutely.
That’s the job of a brand skill. It tells Claude what energy to bring before any words get written. Not just “friendly” or “bold” or whatever limp adjective survived three rebrands. The real stuff: cadence, bite, restraint, humor, visual taste, and what your brand would rather die than sound like.
Do it right, and your content stops reading like a group project between a junior writer, a freelancer, and a chatbot wearing business casual. It starts sounding like one company.
To show how this works, I’ll use a fictional cold brew brand named Hot Take throughout the examples below.
You already have a brand. It’s just scattered everywhere.
Before you write any instructions, go gather all your marketing materials. Get the style guide that you forgot about after the rebrand. Pull the deck that your founder makes every new hire read through. Even dig into those customer support emails that customers reply to, saying thank you for.
The goal here is to round up as much information you’ve got about the current brand voice and herd it into one place for Claude to read.
Save all of this in one working folder. Keep it boring and obvious. Something like:
Claude Brand Skill Source Materials
Then, make subfolders:
01 Brand docs
02 Voice examples
03 Visual examples
04 Content formats
05 Don’t sound like this
For visual guidelines, save screenshots or image files of the brand in the wild. Name the files like a person who wants to find them again:
homepage-hero-example.png
instagram-caption-tone-example.png
bad-example-too-corporate.pdf
Then create one audit document. In that file, make three notes for every asset:
What to keep: The parts that sound right.
What to avoid: The parts that feel off-brand
Why it matters: This is the rule Claude will learn from it.
Here’s an example:
Asset: Q2 product launch email
What to keep: Short sentences, direct CTA, playful first line.
What to avoid: The phrase “seamless experience.” Retire it immediately.
Why it matters: Launch copy should feel confident and useful, not like a SaaS brochure trapped in an elevator.
While you’re in there, you want to be ruthless.
Claude can read a lot, but your job isn’t to throw a junk drawer at it and hope for brand clarity to crawl out like Ripley’s baby in Aliens 4 wearing a tiny crown.
Once you’ve finished the audit, sort the keepers into four piles:
Who the brand is.
How it sounds.
How it looks.
How the voice changes
Those piles become your core skill files:
brand-foundation.md
voice-and-tone.md
visual-guidelines.md
content-formats.md
Hold that thought. This is where the fun starts.
Step 2: Build your foundation so Claude doesn’t have to guess who you are
This is where we build your brand foundation file that keeps all your copy from running into traffic. Call it something basic like brand-foundation.md.
Claude needs to know what kind of brand it is writing for before it starts doing that thing where it gets confident, cheerful, and completely wrong.
Keep this file small. Painfully small. This isn’t the place for every tagline, campaign line, founder quote, abandoned slogan, or sentence that begins with “we empower.” Those belong somewhere else. Possibly a folder. Possibly a bonfire.
The foundation file should include six sections:
Brand summary.
Mission.
Audience.
Positioning.
Personality traits.
What we aren’t.
For the brand summary, write one paragraph. Say what the company does, who it helps, and why I should care.
For Hot Take, the summary would sound more like this:
“Hot Take makes cold brew for people who want it strong, smooth, and a little less precious. No fake wellness glow. Just good caffeine, good flavor, and a morning that doesn’t require waiting behind a lavender mushroom latte with emotional support foam.”
Then write the mission. One sentence. Maybe two if you must. This is the hill the brand would die on.
Not something like this:
“To revolutionize the beverage experience through innovative cold brew solutions.”
For Hot Take, I would keep the mission stupidly clear:
“Make cold brew that tastes good, works fast, and doesn’t make the morning more annoying than it already is.”
Next, define the audience. Claude doesn’t need “women 25-44 with disposable income” as much as it needs the emotional context.
More like this:
“Hot Take is for people who want coffee to do its job. They may be heading into work, to school drop-off, for a client call, for a creative sprint, or just to the emotional obstacle course known as opening Slack.
They like good coffee, but they aren’t trying to make it their whole personality. They don’t want weak coffee, wellness cosplay, or fake premium vibes.”
Then add positioning. This is where you tell Claude how the brand should sit in the market.
“Hot Take lives between gas station coffee and the $9 Starbucks.
It’s better than the sad office fridge situation. Less sensitive than the coffee shop, where ordering requires courage. Strong, smooth, unfussy, and just opinionated enough to feel like someone awake wrote the label.”
Now get to the part everyone loves to skip: personality traits. Pick four or five traits with teeth. Not “innovative,” “customer-obsessed,” or “authentic,” unless you enjoy words that have been left in the sun too long.
For each personality trait, write enough that Claude can make an actual judgment call.
Not just “sharp.” That means nothing. A knife is sharp. So is a bad email from finance.
Write the trait like this instead:
Sharp
Hot Take gets to the point. The copy should have a little bite, but it shouldn’t draw blood. Use shorter sentences. Make the claim clearly. Let headlines carry some edge.
Good: “Strong cold brew. No personality transplant required.”
Too far: “Your current coffee is embarrassing.”
Too flat: “Our cold brew offers a smooth and enjoyable taste experience.”
Do the same for every trait.
Playful
The brand can wink at the reader. It can’t put on a foam finger and start yelling internet slang. A joke is welcome when it makes the line more memorable. A joke isn’t welcome when someone is trying to understand pricing, shipping, refunds, or anything involving their money.
Good: “The 12-pack landed. Your fridge is about to get interesting.”
Too far: “This cold brew slaps harder than your Monday trauma.”
Too flat: “Our 12-pack is now available for purchase.”
Honest
Say what the product does and what it doesn’t do. Tell the truth about the product, that’s enough. This isn’t a spiritual awakening.
Good: “Strong, smooth cold brew for mornings that need backup.”
Too far: “This will change your life.”
Too flat: “Our beverage supports your daily routine.”
Warm
The brand should sound like a person you wouldn’t mind hearing from before 9 a.m. Helpful. Relaxed. Easy to understand. This should feel like waking up to Bill Withers singing Lovely Day. No baby talk. No “Hey bestie.”
Good: “Need help picking a pack? Start with the classic. It plays well with most refrigerators.”
Too far: “Bestie, your caffeine era is calling.”
Too flat: “Multiple pack options are available for customers.”
Steady
Confidence without feeling like a fan girl at a T-Swift concert. This is the trait that keeps copy from spiraling. The brand can be excited.
Good: “New flavor. Same strong cold brew. Available now.”
Too far: “RUN. DO NOT WALK. YOUR MORNING DEPENDS ON THIS.”
Too flat: “We are pleased to announce a new product offering.”
The trick is to give Claude contrast. A good line. A line that goes too far. A line that dies of beige. That’s how the model learns the edges of the voice instead of just memorizing adjectives.
Then write what the brand isn’t.
For Hot Take, I’d write it like this:
Hot Take should never sound smug.
No coffee snob routine. No meme-account-with-a-product energy.
No fake urgency. No fake scarcity. No fake intimacy.
And please, for the love of every tired person in aisle seven, don’t say ‘fuel your journey.’
The “not” list is usually more useful than the “is” list. It catches the moment a brand starts drifting from playful into try-hard, from confident into cocky, from polished into beige.
End the file with a little pre-flight check. Just a few lines Claude can run through before it starts flinging adjectives around.
For Hot Take, that might look like:
Before writing, make sure this sounds like Hot Take on a good day.
Is it useful? Does it feel confident?
Did it get to the point before the reader aged visibly?
Is the joke helping, or is it standing there waving?
Would this sound weird coming from a cold brew brand?
That’s your foundation. Everything else in the skill should connect back to this file. Voice comes from it. Visuals should match it. Content formats should bend around it.
Step 3: Build the voice guide people will want to use
This is the file I would obsess over. Voice is the part people remember when the rest of the page gets chopped up in the sad little gap between meetings.
Voice is the brand on a normal day. Tone is the outfit it wears for the room.
Hot Take can be louder in a product launch email. It should be calmer in a refund email. It can be funny in a social caption. It should probably not be doing stand-up in a shipping delay message.
That’s the point of this file. Claude needs to know the difference between the brand’s personality and the mood of the moment. Then give it examples. Lots of them.
A trait like “honest” doesn’t help much on its own. Honest compared to what? A priest? A mechanic? A brutally direct aunt at Thanksgiving?
Give Claude the before, the after, and why the after works.
After: It’s cold brew. It’s strong. It won’t taste like a compromise.
Now do the same for playful, where the line gets a little more wink and a little less brochure.
Before: Hot Take is now available in a 12-pack.
After: The 12-pack landed. Your fridge is about to get a lot more interesting.
Write five or six of these per trait.
Claude learns a voice the way a new hire does. Your before-and-after pairs are that training ground. The more useful contrast you give Claude, the less it has to guess.
Sadly, my visuals always get treated like they were assembled during a minor hostage situation.
Claude doesn’t need every component state, hover color, and button radius that your product designer has carefully hidden inside Figma. It needs the basics. Things like logos, color palette, fonts, and what not to do.
“Deep blue” means one thing to your designer, another thing to your freelancer, and apparently something very alarming to AI.
Give it the actual codes and the job each color is allowed to do. For instance:
Black: #111111. Use for headlines, body copy, and anything that needs to be easy to read.
Orange: #F45D22. Use for CTAs and callouts. We aren’t painting a traffic cone.
Cream: #FFF8EF. Use for warmer sections or background blocks.
Then add usage rules.
“Use black and cream as the base. Orange is an accent. Aim for 80% neutral, 15% warm background, 5% accent.”
Do the same with type. Don’t just list the font family and call it a day. Tell Claude how the type should feel and how it should behave.
Headlines should feel bold, clean, and editorial.
Body copy should be easy to read.
Use sentence case for headers.
Avoid all caps.
Arial or Helvetica if the primary font is unavailable.
Then get painfully specific about layout. “Generous spacing” sounds nice. It also means nothing. One person’s generous spacing is another person’s abandoned parking lot.
Give Claude something to work with:
Keep the layout rules plain enough that nobody has to interpret them like cave drawings.
Paragraphs should stay short. Two to four lines is plenty. After that, the reader starts looking for an exit.
Each section gets one main CTA. Not three buttons fighting for custody of the click.
For imagery, don’t stop at “natural” or “bright.” That’s mood-board language. Give Claude the taste level.
For Hot Take, the image notes are less “brand guideline” and more “please don’t make this look dead inside.”
“Use real morning stuff. A can on a kitchen counter. A fridge that looks opened by an actual person. Bad lighting? No. Perfect fake lighting? Also no. Somewhere in the middle, where humans live.”
That’s why this file is useful. Claude can help, but only after you show it the taste level. Otherwise, it defaults to safe, and safe is usually just beige wearing better shoes.
And safe is where brands go to become lobby furniture.
Step 5: Teach Claude how the voice changes by channel
A line that works beautifully in an Instagram caption can eat pavement in an investor update. I have seen it happen (no, I didn’t do it, and yes, it was cringe.)
That’s why the content-formats doc (content-formats.md) is created.
This file tells Claude how the brand behaves when the assignment changes. Blog post. Sales email. Support reply. Board deck. Landing page. Social caption. Each one needs its own little rulebook.
For each format, write down:
What the piece needs to do.
How it should be structured.
Which voice traits should be louder.
Which traits should stay in the back seat.
One good example.
One “absolutely not” example.
For Hot Take, here’s an example of social captions:
Turn up: Playful, Sharp, Warm
Turn down: Steady
Keep it short. Let the first line do the work. One joke is plenty. Don’t stack three jokes like a person trying to prove they are fun at a company offsite.
Good: “New 12-pack just dropped. Your fridge has been emotionally preparing for this.”
Too much: “BESTIES. THE 12-PACK ERA IS HERE AND YOUR FRIDGE IS SCREAMING.”
And here’s an example of customer support messages:
Turn up: Warm, Honest, Steady
Turn down: Playful
Be human. Fix the thing. A tiny bit of personality is fine, but nobody wants a stand-up set when their order went missing.
Here’s an example of good: “Sorry about that. Your order should have arrived by now, so we’re checking on it and will make this right.”
And here’s an example of too much: “Uh-oh, looks like your cold brew took a little vacation.”
The goal here is to keep the brand recognizable without forcing it to use the exact same voice everywhere.
Four files down. One to go, and it runs the show. This is called SKILL.md.
This is the one Claude checks first.
Think of it as the little bouncer for the whole skill. It tells Claude what the skill is for, when to use it, which files to read, and what to do before handing you copy that sounds like it was microwaved in a corporate kitchen.
Inside SKILL.md, keep it boring in the places that need to be boring:
Skill name.
Description.
When to use it.
Files to reference.
Workflow.
Checklist.
The description matters more than people think.
Claude uses it to decide when the skill should wake up. So don’t write some vague little fortune cookie like:
“Use this skill to support brand-aligned communication.”
Nobody is typing that. Nobody has ever said that out loud unless there was a webinar involved.
Write the description with the words you use when prompting:
“Use this skill when writing or editing Hot Take blog posts, landing page copy, email campaigns, social captions, product copy, presentation copy, creative briefs, image prompts, or marketing content that needs to match the Hot Take brand voice and visual style.”
That’s the SEO part playing hide and seek. Use the words people search for. Or in this case, the words you prompt with.
Then add the workflow:
First, read brand-foundation.md.
Then check voice-and-tone.md.
If the request involves layout, slides, images, or creative direction, read visual-guidelines.md.
If the request is for a specific format, read content-formats.md.
Before sending anything back, run the checklist.
After that, test it as if you’re trying to break it.
Give Claude a stiff paragraph and ask for a rewrite. Give it a support reply. A homepage section. A social caption. A slide title. A product description. Something boring. Something delicate. Something where the joke absolutely shouldn’t survive.
If the first few drafts are off, don’t keep adding one more to the prompt. That’s how you end up with a final piece that reads like you tried to duct tape your bumper back on.
Open the skill file and find the problem hiding there.Check the stuff you banned. There’s probably a word in there that your team hates but forgot to write down.
Or, maybe you said the brand can be playful, but forgot to mention that refunds aren’t open mic night. Add the line you liked and hated. Then test again.
You aren’t trying to make the prompt longer. You’re trying to make the skill smarter. That’s the whole game.
Teach Claude your voice before it makes more beige
The real magic of a Claude brand skill isn’t that it makes AI faster.
Speed is everywhere. Taste isn’t. Your competitors have speed. The intern with a Canva password and a dangerous amount of confidence has speed.
The win is that your content stops multiplying like generic office wallpaper.
A good brand skill turns Claude from a polite autocomplete machine into a trained creative partner. It knows when to be sharp. When to shut up. When to use the joke. When to keep the joke in its little cage. It understands the words you use, the words you avoid, the rhythm of your sentences, the way your visuals should feel, and the kind of copy that should be escorted off the premises immediately.
For SEOs like me, the job is getting weirder. Now I’m thinking about what happens after the page gets scraped, summarized, cited, and wedged into an answer box with six other brands wearing the same khakis.
The goal is for Joe in San Diego to read one sentence and think, “Yep, that sounds like Anna.” Maybe he laughs. Maybe he hates it. Either way, I made it through the blender.
This doesn’t happen by asking Claude to “make it more on brand,” which is the content equivalent of yelling “be hotter” at a toaster.
So don’t treat the skill like a cute AI side quest. Build the thing as if it were part of your content infrastructure.
Give Claude the good examples. Give it the bad ones, too. Show it the phrases your brand owns, the phrases you never want to see again, and the little judgment calls your best editor makes without thinking.
That’s where the value is. Make sure your brand survives the blender: search results, AI answers, comparison pages, Reddit threads, and the human being skimming all of it while half-watching Netflix.
Creative testing has become a volume game in paid social, but producing more ads doesn’t automatically improve performance. When accounts become flooded with minor variations, budgets fragment, learning phases stretch longer, and performance insights become harder to interpret.
The strongest advertisers today are focusing less on creative quantity and more on differentiated concepts. They’re testing concepts built around audience psychology, emotional resonance, messaging angles, and formats that give algorithms stronger signals to optimize against.
What meaningful creative testing actually looks like
One of the biggest misconceptions about creative testing is that every new asset automatically becomes a fresh test in the algorithm’s eyes. That’s not necessarily true.
Uploading a high volume of ad variations doesn’t automatically create meaningful differentiation. If the only difference between five creatives is the color of the overlay text, the platform can still recognize that the core message, intended audience, and visuals are nearly identical.
Platforms like Meta typically won’t find new audience pockets when this happens, so your creatives compete with one another, leading to delivery overlap. One or two ads may even cannibalize your budget, leaving some variations with little to no impressions.
Meaningful creative testing is rooted in psychology, messaging, emotional triggers, and differentiated creative angles that change how people experience the ad and how algorithms interpret it.
Creative testing works best when concepts truly differ. Lean into different hooks, emotional drivers, positioning, motivations, and formats. That’s where you’ll see meaningful performance shifts.
If creative volume is prioritized too heavily over creative value, it can create performance inefficiencies, waste resources, and add operational drag to your advertising processes.
When your account is flooded with high-volume, low-value creatives, analysis becomes more complicated and pulls you away from higher-level strategic thinking.
Fragmented budgets and longer learning phases
Every time a new asset is introduced, the platform needs data to determine who to show it to, how to optimize delivery, and where it’s most likely to drive results.
When budgets are spread across too many creatives with minor variations, data becomes fragmented, and the algorithm struggles to gather enough conversion signals for each asset to move through the learning phase properly.
Instead of concentrating spend on stronger concepts, your budget becomes diluted across micro-test assets that are unlikely to achieve statistical significance.
Don’t waste budget accumulating inconclusive data that provides little guidance for future creative variations.
The analysis tax
When an account is flooded with assets that contain only minor variations, advertisers get pulled away from macro-level strategy and trapped in the minutiae of the data.
Save yourself time parsing small differences in performance metrics to determine whether the red overlay text outperformed the blue one. Instead, analyze higher-level creative trends.
Misaligned KPIs
While creative production speed and output matter, they shouldn’t be the primary indicators of success. When volume becomes the primary KPI, teams optimize for asset delivery instead of strategic differentiation. There should be a balance between production efficiency and a deeper strategy.
Produce meaningful ads that create measurable impact for both your account and the business.
How to build higher-value creatives
If flooding the system with creative variations that contain only minor tweaks isn’t producing meaningful results, the next question becomes: How do you build high-value creative that actually scales?
Shift away from chasing trends, viral formats, and trendy audio. Instead, use real audience insights from reviews, customer service tickets, social media comments, survey results, and conversations in online forums like Reddit or Quora. Some of the strongest creative inputs already exist within your business.
Look for recurring themes. Are there recurring frustrations, objections, or emotional language patterns? Use AI to analyze them, save time, and uncover messaging insights that will resonate more deeply with your audience.
Once you identify your audience’s vocabulary and pain points, use those findings to shape your messaging and creative concepts. That’s where the real value lies.
High-value creative also doesn’t need high production quality or large budgets. Raw, low-fi content captured on a phone can perform extremely well. I’ve also found that founder-led ad content often performs best because it feels more native and less like polished advertising.
After all, value comes from the message, not the production quality.
Emphasizing creative value doesn’t mean abandoning testing volume entirely. It means sequencing your testing intentionally by using a two-phase framework that separates the pursuit of value from volume.
Phase 1: Macro-testing for value
This initial phase focuses on concept discovery. The goal is to test creative hypotheses.
For example, test three different concepts against one another and identify the winner. Use different formats, emotional angles, and creative styles across those concepts.
Phase 2: Micro-testing for volume
Once you have a clear winner from Phase 1, introduce volume.
If a founder-led video ad delivers a significantly lower CAC and a spike in hook rate, that creates an ideal opportunity to lean into volume. Take that winning creative and iterate on its components to maximize efficiency and extend its shelf life.
Test:
Three different hooks in the first three seconds.
Two variations of pacing or music.
Several CTAs at the end.
By structuring your workflow this way, volume serves to optimize a concept that has already proven valuable.
Shifting from a volume-first approach to a value-first strategy can help pull your organization out of the content mill trap.
Once you implement this process, review your ad accounts each week with these questions in mind:
Are we launching three unique ads or the same ad three times? Review your latest batch of creatives and make sure they contain distinct psychological angles, not just visual changes.
What customer insights drove the last three winning creatives? If you can’t identify the exact objection, review keyword, audience phrase, or use case that inspired a winning ad, you’re scaling on luck and using a strategy that isn’t repeatable.
Is the data telling us a story? Step back from individual metrics and evaluate creative trends holistically. Are videos outperforming images? Are founder-led ads holding attention longer than UGC? Use your metrics systematically to guide your next creative strategy decisions.
Algorithms are powerful and, in many ways, mirror human behavior. They can’t manufacture interest where it doesn’t exist, nor can they turn weak messaging into profits through repetition.
No amount of creative volume will compensate for a lack of strategic value in your ads. Step back, assess the data, identify which concepts are actually working, and give the algorithm something meaningful to learn from to help drive business growth.
B2B PPC advertisers have more options than ever for measuring success. In the past, you had to rely on form-fill data alone. Now, you can feed a wealth of offline conversion data into Google Ads and Microsoft Ads.
It’s tempting to measure every possible metric, but optimizing toward all of them isn’t a good idea. If you’re optimizing toward everything, you’ll probably end up succeeding at nothing.
So how do you know whether you’ve actually driven incremental value, and what are the right success metrics for B2B PPC campaigns? The metrics that matter might not be the ones you’re focused on.
The advertising metrics that matter
I’ve seen advertisers set up offline conversions and get excited because their total conversions increased. Then frustration sets in because they don’t see corresponding increases in their bottom line.
Usually, those conversion increases happen because the advertiser added more conversion actions and set them all to primary. Before making changes, they were only tracking form fills. Afterward, they were tracking form fills, leads, marketing qualified leads (MQLs), sales qualified leads (SQLs), and opportunities.
Instead of one conversion action, they now have four. But the same person could complete all four, meaning leads are being quadruple-counted. A similar issue can happen with platform-reported return on ad spend (ROAS). If you’ve attached conversion values to each action — which you absolutely should do — you’ll also see ROAS increase. Both are false signals created by faulty math.
Focusing only on average cost per action (CPA) can also be misleading. Average CPA can mask your marginal CPA — the cost of acquiring one additional conversion as marketing spend increases. You might be overspending on incremental conversions as you scale your account.
Setting up conversion values is a must for offline conversions. A lot of B2B advertisers get hung up on this step. They say, “We don’t know the value of the conversion at the time it happens. We won’t know that until it works its way through the system.”
While using actual conversion values is ideal, don’t worry if you can’t. Just assign relative values to each conversion action. Here’s a simple example:
In this case, the advertiser is measuring four conversion actions: video views, ungated asset downloads, form fills, and MQLs. MQLs come from offline conversions. The rest are measured through Google Tag conversions.
Each conversion is worth 10x the previous action. Ungated asset downloads are worth 10x a video view, and so on. MQLs are worth 1,000x a video view. The advertiser would rather get one MQL than 999 video views.
If you set arbitrary values, make sure to validate them against real data to ensure they’re directing bidding algorithms accurately. Setting values that are too high for lower-funnel actions can cause the algorithm to favor those easier-to-generate actions while deprioritizing lower-funnel actions.
This happened to us recently with a client who was getting a lot of leads, but very few MQLs and SQLs. We reduced the value of leads by a factor of 10, which made MQLs and SQLs look more valuable to the algorithm and increased MQLs and SQLs relative to leads.
Within two weeks, MQL and SQL volume increased significantly, while leads stayed flat. That might not sound like a good thing, but it was. The client was getting higher-quality leads for the same cost.
By using the right conversion values, even relative ones, you can measure incremental conversion value more effectively.
If you want Smart Bidding to focus on down-funnel actions, you can use campaign-specific goals. You can assign conversion actions at the campaign level, so Smart Bidding only optimizes toward those actions.
You can find the feature in campaign settings in both Google Ads and Microsoft Ads.
Here’s what that looks like in Google:
And here’s what it looks like for Microsoft:
Let’s say you have a campaign that’s driving a lot of leads, but few MQLs and SQLs. You could use campaign-specific goals to optimize only for MQLs and SQLs and ignore leads, even though leads are a primary conversion in the account.
Note: If lower-funnel actions have low volume, this technique may not work. Automation still needs a clear enough signal to optimize toward. So if you only have one or two MQLs in a month, the automation might struggle.
The success metrics for measuring incremental conversions
Looking at simple CPA and ROAS metrics isn’t enough to measure success. You also need to look at each conversion action and measure incremental conversions. A basic way to measure incremental conversions is to establish your baseline first and then measure the CPA and ROAS of conversions at a higher spend level.
For example, let’s say you’re currently spending $5,000 per week and getting an average of 50 conversions, so your CPA is $100.
Now let’s say you increase your weekly spend to $7,500 and end up getting 70 conversions, for an average CPA of $107 — not much higher than the previous CPA of $100.
Your marginal spend is $2,500, marginal conversions are 20, and marginal CPA is $125 — 25% higher than the original CPA.
The difference in CPA may or may not be acceptable for your goals. It might make sense to invest more to increase sales. But you need to understand where your upper limit is. At some point, you may exceed the amount you’re willing to invest for an additional lead while still making fiscal sense.
There are several MMM tools you could use to do this. Some are expensive, and some are low-cost or even free.
For example, Google’s Meridian is an open-source tool and it’s free, but the trade-off is that it requires technical know-how to set up and use.
MMMs also require a significant amount of historical data — two or more years’ worth — but once the data is ingested, they’re fantastic for measuring incrementality.
The true test of whether your success metrics are working
MQLs, SQLs, and closed deals are important. But the true measure of incremental value is revenue or pipeline. That means you need to make sure you’re actually measuring for it.
Not all sales are created equal. You might see one deal for $5,000 and another for $2 million. Both register as closed sales in the CRM, but the two are nothing alike. Which would you rather have?
It’s easy to undervalue conversion actions if you’re using proxy values like the ones described above. Yet pipeline and revenue are often determined outside the 90-day conversion window required for offline conversions.
It’s crucial to look at the data in your CRM and map it back to your paid search campaigns. Are there campaigns or content assets driving relatively few leads and MQLs, but a lot of pipeline? If so, don’t devalue them. Make sure you keep them running and give them enough budget to succeed.
Also, don’t forget about incremental revenue.
If you’re scaling your spend, keep an eye on incremental revenue to find the point of diminishing returns. Doing this can prevent overspending on campaigns or channels once they’re no longer cost-effective.
SEO content is built to rank. Conversion isn’t always the priority when you’re focused on everything else on your SEO checklist.
But as AI Overviews and declining click-through rates make visibility harder to earn, it’s worth thinking about whether your content is built to inspire action once people see it.
TikTok Shop creators have become especially effective at this. The top performers aren’t succeeding because they have massive followings. They’re succeeding because they understand persuasion, consumer psychology, and how to drive action at scale.
You can apply the same principles to written content.
The TikTok Shop conversion formula
In January, I became a TikTok Shop affiliate to better understand what was working on the platform. The more I studied top-performing creators, the more I realized their success wasn’t random.
They were following a formula rooted in consumer psychology. It wasn’t about massive followings or celebrity. On my own videos, 99% of views don’t come from followers.
The creators generating hundreds of thousands of dollars in sales tend to use the same patterns repeatedly, because they understand the psychology that drives action.
The formula usually comes down to four things: visual hooks, psychology levers, storytelling, and relentless testing. Applied to written content, the same structure can drive results.
The foundation: People buy emotionally, then justify rationally
People aren’t rationally evaluating your features and pricing. They’re responding to emotional triggers and rationalizing the decision afterward.
It’s no longer enough to have the right words on your page. They need to connect with the motivations driving someone to act.
The TikTok Shop formula works because it’s built on this principle. Every element in a high-converting video is designed to speak to a natural human desire.
The creators doing this well aren’t necessarily marketing experts. They’re consistently applying consumer psychology.
The 8 human desires in sales psychology
Sales psychology is built around eight natural human desires that motivate behavior:
Care and protection of loved ones.
Survival, enjoyment of life, and life extension.
Enjoyment of food and beverage.
Freedom from fear, pain, and danger.
Sexual companionship.
Comfortable living conditions.
Being superior, winning, and keeping up.
Social approval.
These desires drive human behavior across cultures and contexts. Part of our job as marketers is identifying which ones matter to our audience and weaving them into the content people are already searching for.
Here’s what each desire means and what it might look like in your content.
1. Care and protection of loved ones
The instinct to protect family is one of the strongest levers in consumer psychology. People are more motivated to act when it helps protect the people they love.
What it sounds like in your content:
Insurance: “When something difficult happens, having the right protections in place means your family doesn’t have to think about finances during an already difficult time.”
Home security: “A break-in doesn’t just threaten your belongings. It can shatter your family’s sense of safety. The right security system helps protect both.”
Home repairs: “Fixing a small roof leak today can help protect the home your family depends on tomorrow.”
The shift is subtle but significant. You’re not just describing a product or service. You’re speaking to the person behind the purchase decision and helping them understand the impact of not using your product or service on their family.
2. Survival, enjoyment of life, and life extension
This lever taps into people’s natural desire to protect their health, preserve their quality of life, and make the most of the time they have. It works by showing how a product or decision can help them feel better, stay active, avoid decline, or enjoy life more fully.
What it sounds like in your content:
Supplements: “The right nutrients can help you maintain the energy to keep chasing after your kids, hiking on weekends, or simply feeling like yourself again.”
Outdoor gear: “Reliable gear helps you spend less time worrying about the trail ahead and more time enjoying the experience around you.”
Travel: “Life moves fast. Taking the trip you’ve been putting off for years gives you the chance to experience more of the world while you can.”
3. Enjoyment of food and beverage
People don’t just eat and drink to survive. Food is tied to enjoyment, comfort, celebration, and connection. This lever highlights the emotions and experiences associated with eating and drinking.
What it sounds like in your content:
Meal kit delivery: “Stop treating dinner like a task. Make it the part of your day you actually look forward to with the people who matter most.”
Travel: “From family-owned cafés to local dishes you’ll be talking about long after your trip ends, experience Italy through your taste buds with our food tour.”
Restaurant technology: “When the kitchen runs smoothly, the food gets to the table the way the chef intended.”
4. Freedom from fear, pain, and danger
People are more motivated to avoid pain than pursue gain. This lever applies to almost every industry.
What it sounds like in your content:
Cybersecurity: “One vulnerability could cost you more than your data. Businesses lose [X] every year to attacks that were entirely preventable.”
Domain registration: “Forgetting to renew your domain can mean losing your website, business email, and the online presence you’ve spent years building.”
Financial planning: “The biggest retirement risk isn’t market volatility. It’s reaching retirement and realizing you haven’t saved enough to maintain the lifestyle you worked so hard to build.”
5. Sexual companionship
We’ve all heard the phrase “sex sells” because it taps into a natural human desire. This lever highlights how a product, service, or action can increase confidence, attractiveness, connection, desirability, intimacy, or relationship satisfaction.
What it sounds like in your content:
Fitness and wellness: “When you feel stronger, healthier, and more confident in your own skin, others begin to notice.”
Skincare: “There’s something powerful about leaving the house without makeup, filters, or hesitation.”
Fashion and retail: “The right fit isn’t about impressing everyone. It’s about catching yourself in the mirror and liking what you see.”
People naturally want life to feel easier, more comfortable, and less stressful. This lever highlights how a product improves someone’s environment, routines, convenience, peace, or quality of life.
What it sounds like in your content:
Smart home technology: “Walk into a home that’s already the perfect temperature, the lights are on, and everything is ready for you to relax.”
Headphones: “Whether it’s a noisy commute or a busy household, sometimes comfort means having a quiet moment that’s entirely your own.”
Meal delivery service: “Dinner shouldn’t feel like another task on your to-do list. Enjoy fresh meals without the planning, shopping, or cleanup.”
7. Being superior, winning, and keeping up
People want to feel capable, successful, respected, and ahead of the curve. This lever highlights how a product helps them perform better, achieve more, stand out, or avoid falling behind.
What it sounds like in your content:
SEO agencies or tools: “While competitors are still trying to understand AI search results, you’ll already know where you stand and how to improve visibility.”
Online education: “The skills that are optional today could become the qualifications employers expect tomorrow.”
Fitness: “Most people quit after a few weeks. Consistent training puts you among the small percentage who actually follow through.”
8. Social approval
This is one of the easiest levers to use. It can be as simple as highlighting customer volume, units sold, customer quotes, reviews, or Trustpilot widgets on your site.
You don’t have to tell people your brand has social approval. You need to show it.
In this context, the goal is to highlight your brand’s credibility to customers and Google, not necessarily to convince people that your product gives them social approval.
A bonus layer: Learned desires
Beyond the eight core desires, there are learned desires that can complement the deeper psychological levers. These include:
Convenience (“this takes 10 minutes, not an hour”).
Curiosity (“here’s what most [audience] gets wrong about this”).
Financial gain (“here’s what this is actually costing you right now”).
How this framework improved our results
My team has applied this framework to blog content over the past year. We’re not doing anything dramatic. We’re being more deliberate about which levers we use, for which audience, and where they appear in the article.
While other factors certainly contributed, psychology-based content optimization became a major focus for my junior team. Since July 2025, that focus has driven a 136% increase in total blog visits and a 286% increase in participation order volume from our blog.
Average ranking position across roughly 600 articles has moved from the bottom of page two to the top of page one. No link building. No major technical overhaul. Just really strong content.
The article converting best for us right now, “I bought a domain, now what?”, combines multiple levers throughout the piece. We use:
Authority layering through SME quotes.
Pain-avoidance framing around domain expiration.
Social approval positioning around email credibility.
Building a psychology-driven content strategy
This is easier said than done, and it takes time and patience to implement. But this human layer will become increasingly important for conversion as views and clicks continue declining.
The traffic you do get will likely be more qualified. The question becomes how to increase the likelihood that people take action once they arrive.
Here are three ways to start optimizing your content for conversion.
Understand the person behind the search
Knowing your audience’s demographics isn’t enough. You need to understand:
Their environment.
What they do and don’t have access to.
What influences their decisions beyond your product category.
A small business owner in a large city like New York has different foundational needs than one in a market like West Virginia, where starting a business is less common.
Map levers and intended actions to the journey
Think of this as user journey mapping 2.0. For each page, persona, and journey stage, identify which levers make the most sense and what action you want someone to take. Not every page is designed to sell.
Ask yourself:
What is this person afraid of at this stage?
What do they want to feel after reading this?
What action do they need to take?
Start optimizing
This should become an ongoing part of your content strategy, so start where you can, even if it’s just one page.
Apply one or two levers throughout the page and measure what happens. When something works, reverse-engineer the combination:
Which desire was involved?
Which persona?
Which stage of the journey?
Then expand those insights across related content once you’ve identified a combination that works.
How to make these levers feel natural
A lot of this is the art side of marketing, which means there’s room for interpretation and experimentation.
Use data: You don’t need proprietary research. The Census Bureau, Bureau of Labor Statistics, industry reports, and competitor review platforms are full of data that can make abstract concepts feel tangible. The goal is to use a few well-placed statistics to make a point land.
Name your audience: Users are adding more context to long-tail searches, so reflect those audiences in your content. If you want to reach small business owners, say “small business owners” in your content. If you’re speaking to CMOs, say that. You can address multiple audiences throughout a piece. Just make sure the storytelling still feels cohesive.
Don’t overdo it: Just like too much salt can ruin a meal, overly hyped content can feel inauthentic and turn people away. Keep it simple. Look for one or two strong opportunities to apply these tactics and use them deliberately.
To make this more practical, I’ve built two resources you can use in your process.
A custom GPT for blog-level analysis: Paste in an existing article and ask it to identify which psychology levers are present, which are missing, and where they could be added.
A user journey mapping template: A prebuilt template for mapping personas, desired actions, psychology levers, and content by funnel stage. Copy it and adapt it to your site.
The human layer behind content performance
Technical optimization will continue to play a major role in content visibility. But as visibility becomes harder to earn, the content that performs best will be the content that motivates people to act.
Content that understands the person behind the query — including their fears, motivations, and decisions — will drive stronger business outcomes for SEO and content teams.
TikTok creators and UGC affiliates have already figured this out. They understand how to create persuasive content at scale by tapping into human psychology. You can apply the same thinking to written content.
Frequently asked questions (FAQs) used to sit quietly on support pages and product hubs. Now they influence visibility across AI Overviews, People Also Ask results, and AI search engines that prioritize direct answers to user questions.
More than 80% of AI Overview queries are informational, and 82% have average monthly search volumes under 1,000, Semrush research found. That means longer-tail, lower-volume queries are increasingly driving AI visibility opportunities.
As search behavior becomes more conversational, the quality of your FAQ strategy increasingly depends on the quality of the questions behind it.
The problem is that many brands still rely on the same limited sources for FAQ ideation, even as search behavior evolves. The strongest FAQ opportunities often come from the places where customers are already asking questions naturally — across search, support channels, communities, and AI platforms.
Here are five sources to help you find and prioritize better FAQ content opportunities.
1. Google Search Console data
To some people, this may seem obvious. But it’s also easily overlooked.
Before you begin ideating new FAQ content, audit what’s already gaining traction. Google Search Console (GSC) is the perfect place to do that. It’s an underutilized FAQ research tool because many SEOs filter for their highest-impression or highest-clicked queries, not their highest-intent ones.
To zero in on intent queries, start by filtering your query data for question-based search patterns using this regex:
Then, cross-reference average position against CTR. For queries ranking from the middle of page one through the bottom of page two with low CTR, you’ve likely found a candidate for dedicated FAQ content.
If you rank too low, you may not be relevant enough. If you already rank highly, you probably don’t want to disrupt what’s working. Positions 4-20 can be a sweet spot, although it’s always worth looking outside that range just in case.
To expand from there, use this regex to filter for long-tail queries with eight or more words:
^(\S+\s+){8,}\S+$
If eight-plus-word queries don’t produce enough results, try lowering the threshold to five to seven words. This can be another GSC gold mine for uncovering FAQ opportunities from existing traffic.
It can also turn up a ton of great queries worth tracking in your AI visibility software.
The People Also Ask SERP feature is a valuable signal for understanding how your audience’s questions relate, cluster, and evolve. It’s also one of the few publicly available windows into Google’s understanding of search intent relationships.
Some PAA questions may warrant an entire page, depending on the topic and the value you can provide by going deeper. But they’re also a strong way to identify related FAQs that can strengthen pages you already have.
Tools like AnswerThePublic and AlsoAsked surface these branching question trees at scale, helping you map how topics are explored across search behavior.
There’s still value in doing manual SERP research, too. Search your target topics on Google and expand PAA results across five to 10 priority terms. Look for recurring questions that are relevant to your customers.
Those recurring questions are high-signal candidates that reflect broad informational intent. They’re also more likely to earn AI citations because they represent genuine, repeated user demand.
Exploding Topics doesn’t specifically surface PAA queries, but it’s another valuable research tool for spotting topics that are gaining search interest before they peak. That gives you an opportunity to build FAQ content ahead of demand, anticipating the questions your audience may soon start asking more frequently.
3. Customer-facing teams and internal data
The richest source of FAQ ideation is your own data.
Your customer service team fields real questions from real people every single day. Those questions reveal genuine confusion, hesitation, what gets people excited, what pushes them toward conversion, and the specific language your audience uses.
That matters even more now because AI language models are trained to understand natural, conversational language. Yet somehow, this is one of the data sources we rarely receive when working with agency clients.
Getting this kind of direct customer insight isn’t difficult. It can be as simple as creating a communication channel between teams using tools you already have:
Build a Google Doc where customer service teams can drop in questions.
Create a Slack or Teams channel dedicated to customer questions and concerns.
Periodically ask customer service teams for the top or newest questions they’ve received.
If calls are recorded, and your company’s AI policy allows it, use AI to analyze those calls and surface recurring themes and questions.
Internal site search data is also valuable and similarly underused. If your site has a search function, the queries people type into it are direct signals of intent that weren’t fully satisfied by your existing content, either because the information was difficult to find or didn’t exist.
Pull your site search queries monthly and filter for question-based phrasing or longer queries. These are especially useful for FAQ content on product, service, and support pages where questions tend to be highly specific.
Together, these internal sources can produce FAQ content that performs well at the bottom of the funnel, where specificity and accuracy can significantly impact search visibility, user experience, and conversions.
Unlike keyword tools that aggregate search behavior, Reddit shows you exactly how people phrase their questions, feedback, and opinions.
To start your research, search topics related to your products or services on RedditAnswers or Google to identify the subreddits where your audience gathers. Then search within those communities using your target keywords.
Sort results by Best, Top, and New, and pay close attention to the questions people ask in those threads. You’ll likely uncover topics relevant to your audience that you can incorporate into your content strategy.
Reddit threads are also valuable because they often surface follow-up questions after an initial answer. Those discussions reveal the “OK, but what about …” layer of audience curiosity that PAA data and keyword tools often miss.
Building FAQ content that addresses those secondary questions can improve topical depth and strengthen the topical authority signals that AI models tend to favor.
One of the newest signals for understanding what questions your audience is asking comes from prompt volume data within AI platforms themselves.
The datasets are still imperfect, much like traditional search data. But AI visibility tools like Writesonic and Profound do surface aggregated data on the prompts users enter into AI tools. That gives you insight into questions people may be asking before they turn to traditional search.
Because AI searches tend to be longer, more conversational, and more specific than traditional Google searches, prompt volume data can uncover FAQ opportunities that keyword research tools may miss.
It’s important to remember that FAQ creation isn’t a one-time activity. The questions your audience asks will continue to evolve as your products change, competitors evolve, and the search environment shifts.
For example, a company that sells phone cases will need updated FAQs each time Apple announces a new iPhone. A SaaS brand will need to revisit FAQ content with every product update. A tax software brand has clear triggers for new FAQ content, such as new tax laws or updated forms. A plumber, by contrast, operates in a more established field and may not need to update FAQ content as frequently.
Find a cadence that makes sense for your business, and revisit your research and content regularly. The brands winning search visibility aren’t the ones that created strategic FAQ content once. They’re the ones that keep showing up with the right answers as customer questions evolve.
Getting cited in AI answers is becoming a common visibility metric. But citations alone don’t explain why certain brands consistently appear in ChatGPT, Google AI Mode, Perplexity, and other AI search systems.
Citations reflect visibility outcomes, not the underlying systems that produce them. AI platforms prioritize brands with strong semantic presence across training data, reviews, media coverage, search systems, and interconnected web entities.
That’s why GEO is really two visibility challenges happening at once: building long-term brand weight inside AI systems while also creating content that survives modern retrieval pipelines.
AI recommendations are shaped during both retrieval and synthesis. Brand depth is what increases your odds in both systems.
GEO means playing two games at once
Each layer influences visibility differently.
Game 1: Parametric weight
Brands act as coordinates in an LLM’s embedding space, defined by the density and consistency of signals in training data.
This parametric weight is built slowly over months and years through consistent presence across the web. If messaging is inconsistent, the brand’s vector becomes fuzzy, reducing recall and confidence.
A brand with little parametric weight is functional, forgettable, and interchangeable. You can’t easily alter what a model has already internalized during training, so most efforts are directed toward future training cycles.
Focusing exclusively on citations for months means neglecting the structural foundation that eventually makes those citations unavoidable.
When a system like Google AI Mode or ChatGPT Search fires its retrieval pipeline, does your content make it through?
About 85% of brand mentions in AI search come from external domains, not the brand’s own site. Every major AI search system starts with retrieval, but each handles it differently:
Perplexity retrieves, ranks, and embeds citations into the context window before the LLM generates a single token. The model synthesizes answers from retrieved evidence rather than directly from training data.
Google AI Mode decomposes a single query into 8-12 parallel subqueries across the live web, Google’s Knowledge Graph, and specialized data sources before synthesizing a response. Google calls this query fan-out.
ChatGPT search expands a query into five or six semantic variations, retrieves 35 to 42 candidate URLs, disqualifies 83% before extraction, and synthesizes three to five citations in the final response. Retrieval is typically skipped only for nonfactual prompts, such as creative writing or basic math.
In fan-out systems, you compete across 8-12 parallel subqueries simultaneously.
Citations are receipts
Only 6% to 27% of frequently mentioned brands are also top-cited sources. Models can know a brand without citing it.
Citation frequency tracks output presence, not the retrieval and synthesis decisions that surfaced the brand in the first place. Optimizing for citations focuses on the receipt rather than the underlying driver.
Brand depth, built through density, consistency, and cross-source coverage, is what makes a brand the statistically low-risk answer before a citation is ever generated.
Brand depth: How human brains and LLMs default to the familiar
The human brain operates similarly to LLMs. We manage a massive volume of daily decisions by relying on mental frameworks and heuristics that have been built over time.
This idea is rooted in predictive processing theory, which describes the brain as a forecasting engine that uses past information to minimize errors.
LLMs and human cognition handle ambiguity in similar ways: Both prioritize information that is most densely established within their respective systems.
Brand element
Human brain
LLM
Memory and recall
Episodic and emotional, triggered by sensory cues.
Statistical frequency and co-occurrence density in training data. High frequency increases recall.
Brand identity
Sensory and visual: logo, typography, and packaging.
Semantic proximity: adjectives, reviews, and articles associated with the brand name. A coordinate in embedding space.
Building trust
Social proof, word-of-mouth, and personal trial.
Parametric authority: training data weighted toward high-authority sources.
Handling mistakes
Forgiveness through empathy. An apology can repair the relationship.
Data permanence: models consolidate patterns, not intent. Negative signal floods persist until newer data outweighs them.
The recommendation
Impulsive and bias-driven: scarcity, FOMO, and halo effect.
Synthesis-weighted: shaped by what’s most densely represented in parametric memory and retrieved sources simultaneously.
Getting technical about branding with brand depth
AI models and Google’s Knowledge Graph learn from many of the same trusted websites. AI models learn by identifying which words frequently appear together, while Google uses that same information to build a network of connected facts.
Google’s systems specifically evaluate entity salience, entity coherence, and inter-entity relationship density.
Entity salience
How prominent and distinct your brand is within a specific topic cluster. Entity salience influences citation probability.
Google asks: How prominent is this brand within a topic cluster?
LLMs ask a similar question at inference time: Which entities have enough statistical weight to surface when a topic is queried?
Low salience means you’re retrievable only through exact branded queries. High salience means you appear when the topic comes up, not just when your name is searched.
The consistency of your brand’s identity across all retrieved contexts.
Inconsistent naming, conflicting positioning, and contradictory dates signal that an entity is unreliable. LLMs trained on that same corpus learn a fragmented, low-confidence representation.
The model fills gaps created by entity incoherence, leading to brand drift, where the model’s version of your brand slowly diverges from reality because the training signal was never stable enough to anchor it.
Inter-entity relationship density
The strength and number of connections between your brand and other authoritative entities, including products, concepts, and proofs.
Inter-entity relationship density influences associative retrieval paths.
In agentic systems like Deep Research, AI Mode, and Perplexity Pro, each reasoning step is a retrieval event. Relationship density determines whether your brand survives hop two and hop three.
A brand that only exists at the center of its own graph disappears the moment the query moves one step sideways. GlobalLinkInfo and LatentEntity in Google’s Content Warehouse map these inter-entity edges.
The RAG layer is where site quality becomes a gate
Mark Williams-Cook documented a site quality score in December 2024. The score uses a 0-to-1 scale, and sites scoring below roughly 0.4 are not retrieved as candidates, regardless of optimization efforts.
That matters because retrieval eligibility influences which entities and sources repeatedly enter AI systems in the first place. Brand integrity becomes an infrastructure problem. You can’t optimize your way into LLM citations if you haven’t first built the entity coherence and relationship density that make your brand consistently retrievable.
The more co-occurrences you have, the higher your mutual information score, and the more often you appear in answers.
Clinique’s Black Honey lipstick is a good example of how this works in practice because of its strong entity depth:
Concept: Co-occurs with “universally flattering” and “my lips but better” (MLBB) value propositions.
Trend: Co-occurs with “TikTok virality,” driven by high velocity in 2021.
Competitors and dupes: Co-occurs with “e.l.f. Black Cherry dupe,” reinforcing benchmark status.
Proof: Co-occurs with “Liv Tyler” and “Arwen,” creating a cultural anchor.
History: Co-occurs with “1971,” reinforcing longevity.
Because of this density, AI systems repeatedly surface Black Honey when answering questions about universally flattering lipstick.
High recall: AI models are more likely to recall and mention Clinique Black Honey across a wide range of relevant queries, including “best universally flattering lipsticks,” “viral makeup trends,” and “iconic ’90s beauty.”
High authority: The depth of co-occurrence, including historical evidence, cultural context, and product variants, provides AI models with sufficient information to generate detailed, authoritative, and multifaceted answers.
Building for retrieval, recall, and recommendation
Preference is what survives. Build for the layer that determines synthesis weight and for what happens inside the retrieval funnel.
When your brand is specific, consistent, and densely connected across topical clusters, it becomes easier for AI systems to retrieve, synthesize, and recommend.
Focus on what survives the retrieval funnel
Specific, data-rich, hard-to-reproduce content gets retrieved and cited. Academic literature refers to this as adaptive retrieval.
Generic, predictable content gets skipped because the model can generate it on its own.
Low entropy gets ignored
High entropy gets cited
“Our coffee is smooth and delicious.”
“The Gesha variety from Hacienda La Esmeralda in Boquete, Panama. Grown at 1,700 meters. Water at 94 C. Brew ratio 1:16.”
The second version anchors named entities, including a variety, an organization, a location, and quantitative values. These are details the model can’t plausibly generate without a source.
Actionable tip: Add high-density assets, including company history, team bios, and ISO certifications, designed to serve as grounding data for retrieval-augmented generation (RAG) systems.
Build AI navigation maps
Your website functions like a knowledge graph. AI systems use internal links to build a semantic map of your domain.
Embed links that define logical relationships between entities and create clear paths for crawlers to follow. Structure links around the user’s decision journey, which often mirrors AI retrieval paths:
Topic → Subtopic (broad context)
Subtopic → Product (specific solution)
Product → Review (social proof)
Review → Return policy (trust signal)
Return policy → Organization (entity credibility)
Avoid orphan pages
Pages with no meaningful incoming anchors are likely demoted in processing. They don’t accumulate siteAuthority or NavBoost signals.
The fix is to give these pages strategic internal links that connect them to the graph, or delete them. If a page isn’t worth linking to, is it worth human or bot attention?
Citation frequency studies are symptom trackers, not diagnostic tools. They can tell you that certain brands appear more often. They can’t reliably explain whether that visibility comes from training data, RAG retrieval, entity salience, or category dominance.
Build the thing that causes citations, not the thing that imitates them.
There are two questions that come up in at least one meeting per day for every single business.
“How do we get back our Google clicks?”
“How do we show up in all the LLMs?”
The answer to both of these questions is one that very few want to hear. Hold your breath…
Build your brand. Gone are the days of “add some more keywords” and “how many links would it take to…”
Does this mean search-and-answer bots can’t be manipulated? hell no, but the likelihood that these “tactics” will deliver long-term, consistent value is nil.
Ever heard of Crayola? Probably. It is only a little ~$1 billion company, and the default answer most people think of when someone says crayons.
How about Monday Mandala? It is owned by retired school teacher Inez Stanaway and focuses heavily on coloring pages and printable activities.
Now for the fun question: which one do you think drives more traffic for coloring-related searches?
If you guessed Monday Mandala, you’re right.
Google still rewards usefulness, and honestly, that is a good thing. Nobody is going bankrupt because they downloaded coloring pages from the “wrong” website.
But here is where things start to shift.
If I asked 10 people to name a crayon company, how many would say “Crayola”? If I asked those same 10 people to name a coloring-page website, how many would say Monday Mandala?
Monday Mandala won clicks.
Crayola won memory.
And in a world of AI results, recommendations, and answer engines, recognition starts to matter again because traffic gets clicks, but branding compounds and extends well beyond a SERP change or algorithm update.
Search fragmented. Brand didn’t.
Search used to be simple.
Someone had a question, opened Google, clicked a result, and landed on your website. Success was largely measured by clicks, traffic, and actions users completed once on the site.
Many people, including business owners, think they are entitled to “free clicks”. Wake up, Google doesn’t owe you.
The world of building a business off search traffic still exists (I mean, look at MM above), but it is no longer the only game in town. It is also a bigger risk today to rely on search traffic than it ever was in the past.
Today, answers happen everywhere. Google AI Overviews, ChatGPT, Perplexity, Reddit threads, Slack and Teams conversations, LinkedIn posts, YouTube videos… users no longer search exclusively within Google’s ecosystem. Traditional search is now just one piece of a much larger answer ecosystem.
So what survives when users never click?
Brand memory
People remember names they have seen before. They remember positive experiences, friends’ recommendations, repeated exposure, and trusted companies.
Nobody remembers your title tag. (Ooof… that one hurt me a little, too.)
This is exactly why the brand suddenly feels like the answer to every executive question about SEO, AI, video, etc. When users search for answers across platforms, what travels with them is not your website.
It is your reputation.
Not domain authority. Not inbound links. Not keyword density. Not a note from your mom saying you’re a good kid, I mean…and not your Reddit karma score.
Proof that you can rank pages, gain traffic, and ride algorithm shifts to capture massive spikes in visibility. Some businesses will continue to grow primarily through search traffic for years to come, but…
Visibility still matters because people need to find you. Recognition matters because they need to remember you after they leave. Trust matters because recommendations and reputation increasingly influence buying decisions. Proof matters because people want validation beyond your own marketing claims.
Finally, there is brand presence.
Do you exist outside your website?
Newsletters, LinkedIn, PR, podcasts, communities, reviews, mentions, and brand searches are no longer “nice to haves to appease the bots.”
These are signals that users, search engines, and AI systems use to understand who deserves attention.
Your homepage is no longer your brand. It’s just one piece of evidence supporting it.
The uncomfortable truth
After posting about a tool I like on LinkedIn, someone reached out, surprised I would support it. The company was getting blowback over programmatic SEO, and there were rumors that some larger brands using it tactically had been penalized by Google.
My response was simple:
People kill people with Teslas, too.
Tools are not the problem
People can use AI to create junk. They can use SEO to manipulate rankings. They can manufacture links, game prompts, and chase short-term visibility spikes. (hello listicles!)
Login
