Ads are now being tested in ChatGPT in the U.S., appearing for some users across different account types. For the first time, advertising is entering an AI answer environment β and that changes the rules for marketers.
Weβve used AI as part of ad creation or planning for years across Google, LinkedIn, and paid social. But placing ads inside an AI system that people trust to help them think, decide, and act is fundamentally different. This is not just another channel to plug into an existing media plan.
The biggest question is not targeting. Itβs psychology. If advertisers simply replicate what works in search or social, performance will disappoint, and trust may suffer.
To succeed, brands need to understand how and why people use ChatGPT in the first place and what that means for attention, relevance, and the customer journey.
ChatGPT is a task environment, not a feed
People open ChatGPT to do something. That might be:
Solving a specific problem.
Refining a shortlist.
Planning a trip.
Writing something.
Making sense of a complex decision.Β
This is very different from feed-based platforms, where people expect to scroll, be interrupted, and discover content passively.
In task-based environments like ChatGPT, behavior changes:
Goal shielding: Attention narrows to completing the task, filtering out anything that does not help progress.
Interruption aversion: Unexpected distractions feel more irritating when someone is focused.
Tunnel focus: Users prioritize clarity, speed, and momentum over exploration.
This is why clicks are likely to be harder to earn than many advertisers expect. If an ad does not help the user move forward with what they are trying to achieve, it will feel irrelevant, even if it is topically related.
Add to this the fact that trust in AI environments is still forming, and the tolerance for poor or interruptive advertising becomes even lower.
When there are no search volumes, behavior becomes the strategy
For years, search volume has shaped how we plan.
Keywords told us what people wanted, how often they wanted it, and how competitive demand was. That logic underpinned both SEO and paid media strategy.
ChatGPT changes that.
People are not searching for keywords. They are outsourcing thinking. They describe situations, ask layered questions, and seek outcomes rather than information alone.
There is no query data to optimize against. Instead, success depends on understanding:
What job the user is trying to get done.
Which part of the journey they are choosing to outsource to AI.
What kind of help they need in that moment.
This is where behavioral insight replaces keyword demand as the strategic foundation.
From keyword intent to behavior mode targeting
Rather than planning around queries, advertisers need to plan around behavior modes, the mindset a user is in when they turn to ChatGPT.Β
A useful way to think about this is:
Explore mode: The user is shaping a perspective or seeking inspiration.
Ads that work here help people start, offering ideas, options, or reframing the problem.
Reduce mode: The user is simplifying and narrowing choices. Effective ads reduce effort by clarifying differences and highlighting relevant trade-offs.
Confirm mode: The user is looking for reassurance. This is where trust matters most: proof, reviews, guarantees, and credible signals.
Act mode:The user wants to complete the task. Ads that remove friction perform best, clear pricing, availability, delivery, and next steps.
These modes closely mirror the human drivers we already recognize in search behavior: shaping perspective, informing, reassuring, and simplifying.
The difference is that ChatGPT compresses these moments into a single interface.
A key shift advertisers need to internalize is that relevance in ChatGPT is not about being related. It is about being useful.
An ad can be perfectly aligned to a category and still fail if it does not help the user complete their task.
In a task environment, anything that creates extra work or pulls attention away from the goal feels like friction. This means the creative rules change.
High-performing ads are likely to behave less like traditional advertising and more like:
Tools.
Templates.
Guides.
Checklists.
Shortcuts.
Decision aids.
They fit into the flow of what the user is doing.
Generic brand ads, pure awareness messaging, and content that feels like a detour are likely to underperform.
Helpful content becomes the bridge across channels
The same assets that make a strong ChatGPT ad β practical guides, frameworks, calculators, explainers, and reassurance-led content β also do much more than support paid performance.Β
They build authority for SEO and generative optimization, earn coverage and credibility through digital PR, and reinforce brand trust across social and owned channels.
This is where silos start to break performance.
Paid media teams cannot create βhelpful adsβ in isolation if SEO teams are working on authority, PR teams are building trust signals, and brand teams are shaping voice independently. In AI-led discovery, these signals converge.
The most effective ads may borrow from:
Brand voice for clarity and consistency.
Trusted voice through reviews, experts, or third-party validation.
Amplified voice via media coverage and recognizable authority.
The line between advertising, content, and credibility becomes increasingly blurred.
Measurement needs a reset
Judging ChatGPT ads purely on click-through rate risks missing their real impact.
In many cases, these ads may influence decisions without triggering an immediate click. They may help a brand enter a shortlist, feel safer, or be remembered when the user returns later through another channel.
More meaningful indicators may include:
Shortlist inclusion.
Brand recall.
Assisted conversions.
Branded search uplift.
Direct traffic uplift.
Downstream conversion lift.
This reinforces the need for teams to work more closely together. If performance is distributed across the journey, measurement and accountability must be too.
This is not simply a new ad format. We are looking at a behavioral shift.
The brands most likely to succeed will not be the ones that move fastest or spend the most. They will be the ones who understand:
What people actually use ChatGPT for.
Which moments of the journey are being outsourced to AI.
How to support those moments without breaking trust.
A practical starting point is returning to jobs-to-be-done thinking. Map the actions that happen before someone buys, inquires, or commits and identify where AI reduces effort, uncertainty, or complexity.
From there, the question becomes more powerful than βhow do we advertise here?β:
How can we be genuinely helpful at the moment it matters?
That mindset will not only shape performance in ChatGPT, but across the wider future of AI-led discovery. And in that world, behavioral intent will matter far more than keywords ever did.
Custom GPTs can help SEO teams move faster by turning repeatable tasks into structured workflows.
If you donβt have access to paid ChatGPT, you can still use these prompts as standalone references by copying them into your notes for future reuse. You will need to tweak them for your teamβs specific use cases, because they are intended as a starting point.
Working with AI is largely trial and error. To get better at writing prompts, practice with small tasks first, iterate on prompts, and take notes on what gets you good outputs.Β
AI also tends to ramble, so it helps to give strict guidelines for formatting and to specify what not to do. You can upload resources and articles to follow and provide clear context, such as defining the role and audience upfront.
The seven prompts below are designed to help you start building custom GPTs for planning, analysis, and ongoing SEO work.
1. Project plan GPT
Using past examples of project plans, create a GPT that will help you make a draft for this yearβs focus areas.
How to set it up
Input project plans from previous years.
Give it a specific format to follow.
Consider how many items or sections to include.
Add specific details based on you or your team.
(Optional) Copy notes and feedback from your team or retrospective.
Example prompt
Based on last yearβs project plan, make my project plan for this year. Here are the focus areas and problem areas to include.
Give me a bulleted list with the three most important items for me (or my team) to focus on for each quarter of this year. At least one item should cover link building.
Include a one-sentence summary of why you recommend each item and at least two KPIs to measure success.
[Insert last yearβs plan.]
Now poke holes in your plan. Give me three reasons I should not focus on these items based on the risks. Include sources for your notes.
Hook up your performance dashboards or custom GA reports to ChatGPT and let it do the initial legwork in identifying issues. Then make a list of items to investigate yourself.
How to set it up
Connect your reporting tools or upload reports directly.
Give specific direction for what to look for.
Include the cadence you want to look at, like a daily or weekly report.
Give examples of types of pages or categories to compare.
Example prompt
Here is the weekly site report. Give me your analysis of how the site performed compared to last week. Include a three-sentence summary of the sessions, conversions, and engagement.
List three wins and three misses in bullet format. Color-code each item based on how good or bad each item is.
[Insert report doc.]
3. Competitor analysis GPT
Check out whatβs working and whatβs not on competitor sites and get insights for yours. Itβs most helpful to connect to a tool like Semrush or Ahrefs.Β
How to set it up
Connect tools like Ahrefs or Semrush, or upload a report.
Identify competitors to analyze and top pages and folders.
List key metrics to compare.
Set up unique prompts for page, keyword/topic, folder, and domain-level comparison.
(Optional) Create documentation on identifying which metrics to dig deeper.
Example prompt
You are an SEO analyst performing competitor analysis to identify areas to improve your website. Check out these URLs and compare them. Give me a table with each URL in the rows and these columns: backlinks, average rank, top keyword, sessions, and estimated value.
Below that, give a two-sentence summary of who wins in each category and why. Use the criteria in this link to make your judgments, citing sources for each.
AI has gotten much better over the last few months at analyzing images. Plug in SERP screenshots from your own searches and compare it to a web search from the GPT. Build this into a competitive SERP landscape analysis to see things like who appears in both searches vs. only one.
How to set it up
Identify search results and keywords to compare.
Take screenshots in incognito mode for comparison.
Example prompt
Do a web search for [your keyword here]. Show me what you are seeing in the search results.
Compare it with this screenshot and list the differences. Then include a bulleted list of what the results seen most often have in common.
Turn your design or UX teamβs resources into an easy-to-use helper. This is especially helpful for editorial teams that do not want to search through endless documentation for quick advice.
How to set it up
Upload your teamβs documentation or your favorite UX articles.
Find pages with poor bounce or engagement stats.
Integrate the tool into standard page updates.
Example prompt
You are an SEO writer working on improving user engagement. Open this page. Check to make sure it follows all of our design rules.
List each violation, along with a source, explaining what is wrong and what to do instead. Then check to see whether there are any relevant page template patterns from the brand book that could apply to this type of page.
6. Tech SEO check GPT
Set up a daily or weekly tech SEO check to do the bulk of the analysis for you.Β
How to set it up
Connect any tools like Google Search Console, or upload reports.
List the top metrics to check, like Core Web Vitals, page speed, and console errors.
Identify top pages to run a more comprehensive check.
Set up reminders to run it daily or weekly, or connect it to Slack to export results directly.
Example prompt
Based on the latest CWV report, identify problem pages that need a speed improvement audit. Create the list in a table, with the URLs in rows and columns for speed, issues identified, and suggested fixes. Make a separate list of pages that have improved, along with the actual scores.
While ChatGPT cannot directly create slides yet without an add-on or third-party connector, it can create the content for you to paste into your slides. Combine it with your performance, testing, tech SEO, and competitor GPTs for a well-rounded summary of overall site status with relevant context.
How to set it up
Gather data from your other GPTs.
Choose the ones to present.
(Optional) Upload past presentations for reference.
Example prompt
Pretend you are setting up a slide deck. The audience is other members of the SEO team. Format this summary from my Performance GPT into a slide.
Give me a header, subheader, and key bullets and takeaways. The tone should be straightforward but professional. Limit bullets to one line. Round all numbers to zero decimals. Suggest three examples of imagery and graphics to use.
Custom GPTs are most useful when they sit alongside the tools and processes SEO teams already use. Rather than replacing dashboards, audits, or documentation, they can handle first passes, surface patterns, and standardize how work gets reviewed before a human steps in.
Used this way, the prompts in this article are less about automation for its own sake and more about reducing friction in common SEO tasks, from planning and reporting to SERP analysis and technical checks.
Microsoft's financial earnings report reveals that OpenAI accounts for 45% of its Azure cloud computing backlog, yet Microsoft's net income jumped by $7.6 billion, primarily due to its investment in OpenAI.
Login
TechRadar - All the latest technology news
Latest from Windows Central
