As the SaaS market reels from a sell-off sparked by autonomous AI agents like Claude Cowork, new data shows a 53% drop in AI-driven discovery sessions. Wall Street dubbed it the “SaaSpocalypse.”

Whether AI agents will replace SaaS products is a bigger question than this dataset can answer. But the panic is already distorting interpretation, and this data cuts through the noise to show what SEO teams should actually watch.

Copilot went from 0.3% to 9.6% of SaaS AI traffic in 14 months

From November 2024 to December 2025, SaaS sites logged 774,331 LLM sessions. ChatGPT drove 82.3% of that traffic, but Copilot’s growth tells a different story:

SaaS AI Traffic by Source (Nov 2024 – Dec 2025)

Source	Sessions	Share
ChatGPT	637,551	82.3%
Copilot	74,625	9.6%
Claude	40,363	5.2%
Gemini	15,759	2.0%
Perplexity	6,033	0.8%

Starting with just 148 sessions in late 2024, Copilot grew more than 20x by May 2025. From May through December, it averaged 3,822 sessions per month, making it the second-largest AI referrer to SaaS sites by year-end 2025.

Investors erased $300 billion from SaaS market caps over fears that AI agents will replace enterprise software. But this data points to a less dramatic force: proximity.

Copilot thrives because it captures intent inside the workflow. Standalone tools saw a 53% traffic drop while workplace-embedded AI grew 20x.

Software evaluation is work, and Copilot sits where that work happens.

When someone asks, “What CRM should we use for a 20-person sales team?” while building a business case in Excel, that moment is captured—one ChatGPT never sees. The May surge reflects that activation: Microsoft 365 users realizing they could research software without opening a new tab.

41.4% of SaaS AI traffic lands on internal search pages

SaaS AI discovery sends users to internal search results first, not product pages.

Top SaaS Landing Pages by LLM Volume

Page Type	LLM Sessions	% of AI Traffic	Penetration vs Site Avg
Search	320,615	41.4%	8.7x
Blog	127,291	16.4%	8.1x
Pricing	40,503	5.2%	3.2x
Product	39,864	5.1%	2.0x
Support	34,599	4.5%	2.1x

Despite capturing 320,615 sessions — more than blog, pricing, and product pages combined — this dominance likely reflects LLM limitations, not superior content. LLMs route users to search when they lack a specific answer.

For SaaS companies watching their stock crater, that’s useful news: there’s a concrete technical fix. The 41.4% isn’t an existential threat. It’s a crawlability problem.

When an LLM can’t find a direct answer, it defaults to the site’s internal search. The AI treats your search bar as a trusted backup, assuming the search schema will generate a relevant page even if a specific product page isn’t indexed.

At 1.22%, search page penetration is 8.7x the site average. The cause is a “safety net” effect, not optimization.

When more specific pages — like Product or Pricing — lack the data an LLM needs, it falls back to broader search results. LLMs recognize the search URL structure and trust it will return something relevant, even if they can’t predict what.

Blog pages follow with 127,291 sessions and 1.13% penetration. These are structured comparison posts — “best CRM for small teams” or “Salesforce alternatives” — that LLMs cite when they have specific recommendations.

Pricing pages show 0.45% penetration; product pages, 0.28%. When users ask about software selection, LLMs route to comparison surfaces — search and blog — first. Direct product or pricing pages get cited only when the query is already vendor-specific.

The July peak and Q4 decline reflect corporate work cycles

SaaS AI traffic peaked in July at 146,512 sessions, then declined steadily through Q4:

Month	Sessions	Change
July 2025	146,512	Peak
August 2025	120,802	-17.5%
September 2025	134,162	+11.1%
October 2025	135,397	+0.9%
November 2025	107,257	-20.8%
December 2025	68,896	-35.8%

Every platform declined. ChatGPT’s volume was cut in half, dropping from 127,510 sessions in July to 56,786 by year-end. Copilot fell from 4,737 to 2,351. Perplexity dropped from 7,475 to 3,752.

Two factors drove the slide:

• People weren’t working. August is vacation season, November includes Thanksgiving, and December is the holidays. Software research happens during work hours; when offices close, discovery drops.
• Q4 ends the fiscal “buying window.” Most teams have spent their annual budgets or are deferring contracts until Q1 funding opens. Even teams still working aren’t evaluating tools because there’s no budget left until the new fiscal year.

The July peak reflects midyear momentum: people are working, and Q3 budgets are still available. The Q4 decline reflects both fewer researchers and fewer active buying cycles.

This is where the sell-off narrative breaks down.

Investors treat a 53% traffic drop as proof that AI discovery is stalling. But the data aligns with standard B2B fiscal cycles.

AI isn’t failing as a discovery channel. It’s settling into the same seasonal rhythms as every other B2B buying behavior.

What this data means for SEO teams

Raw traffic numbers don’t show where to invest. Penetration rates and landing page distribution reveal what matters.

Track penetration by page type, not site-wide averages

SaaS shows 0.41% sitewide AI penetration, but that average hides concentration. Search pages reach 1.22%—8.7x higher. Blog pages hit 1.13%. Pricing pages are at 0.45%. Product pages lag at 0.28%.

If you’re only tracking total AI sessions, you’re measuring the wrong metric. AI traffic could grow 50% while penetration on high-value pages declines. Volume hides what matters: where AI users concentrate when they arrive with intent.

Action:

• Segment AI traffic by page type in GA4 or your analytics platform.
• Track penetration (AI sessions ÷ total sessions) by page category monthly.
• Identify pages with elevated concentration, then optimize those surfaces first.

Search results pages are now a primary discovery surface

Internal search captures 41.4% of SaaS AI traffic. If those results aren’t crawlable, indexable, or structured for comparison, you’re invisible to the largest segment of AI-driven buyers.

Most SaaS sites treat internal search as navigation, not content. Results return paginated lists with minimal product detail, no filter signals in URLs, and JavaScript-rendered content LLMs can’t parse.

Action:

• With 41.4% of traffic hitting internal search, treat your search bar as an API for AI agents.
• Make search pages crawlable (check robots.txt and indexability).
• Add structured data using SoftwareApplication or Product schema.
• Surface comparison data — pricing, key features, user count — directly in results, not just product names.

Make your data legible to LLMs — pricing and content both

The sell-off is pricing in obsolescence, but for most SaaS companies the real risk is invisibility. Pricing pages show 0.45% AI penetration—below the 0.46% cross-industry average. Blog pages captured 127,291 sessions at 1.13% penetration, but only when content directly answered selection queries. The pattern is clear: LLMs cite what they can read and parse. They skip what they can’t.

Many SaaS sites still gate pricing behind contact forms. If pricing requires a sales conversation, AI won’t recommend you for “tools under $100/month” queries. The same applies to blog content. When someone asks, “What CRM should I use?” the LLM looks for posts that compare options, define criteria, and explain tradeoffs. Generic thought leadership on CRM trends doesn’t get cited.

Action:

• Publish pricing on a dedicated, crawlable page. Include representative examples, seat minimums, contract terms, and exclusions.
• Keep pricing transparent. Transparent pages get cited; gated pages don’t.
• Replace generic blog posts with structured comparison pages. Use tables and clear data points.
• Remove fluff. Provide grounding data that lets AI verify compliance and integration capabilities in seconds, not minutes.

Workplace-embedded AI is growing 10x faster than standalone LLMs

Copilot grew 15.89x year over year. Claude grew 7.79x. ChatGPT grew 1.42x. The fastest growth is in tools embedded in existing workflows.

Workplace AI shifts discovery context. In ChatGPT, users are explicitly researching. In Copilot, they’re asking questions mid-task—drafting a proposal, building a comparison spreadsheet, or reviewing vendor options with their team.

Action:

• Track Copilot and Claude referrals separately from ChatGPT. Monitor which pages these sources favor.
• Recognize intent: these users aren’t browsing — they’re mid-task, deeper in evaluation, and closer to a purchase decision.
• Show up in workplace AI discovery to support real-time purchase justification.

Survival favors the findable

The 53% drop from July to December reflects AI usage settling into the software buying process. Buyers are learning which decisions benefit from AI synthesis and which don’t. The remaining traffic is more deliberate, concentrated on complex evaluations where comparison matters.

For SaaS companies, the window for early positioning is closing. The $300 billion sell-off is hitting the sector broadly, but the companies that survive the repricing will be those buyers can find when they ask an AI agent, “Should we renew this contract?”

Teams investing now in transparent pricing, crawlable data, and comparison-focused content are building that findability while competitors debate whether AI discovery matters.