AI Search Insights: Debunking Myths About Rankings, Citations, and Click Impact

| Editorial Team On 20 Aug, 2026

AI Search Myths Shattered: What 15 Million Data Points Reveal About Rankings, Citations, and Clicks

Ahrefs has debunked nine of the most persistent AI search myths using data from 15 million data points across ChatGPT, Google AI Overviews, Gemini, Perplexity, and Copilot — and the findings challenge nearly everything marketers thought they knew.

The study arrives at a pivotal moment. As AI-powered search tools reshape how users discover information online, a flood of unverified tactics has swept through the SEO industry. Ahrefs' research — spanning more than 50 AI studies and reviewed editorially by Jennifer McDonald — cuts through the noise with hard numbers and controlled experiments that expose shortcuts for what they are: guesswork dressed up as strategy.

For marketers still finding their footing with the fundamentals of artificial intelligence and how it works, the scale and speed of these shifts can feel disorienting. This research provides rare empirical grounding in a space where opinion frequently outpaces evidence.

The Myths That Cost Marketers the Most

"Best-of" Lists Help Your Brand — But Mostly Your Competitors

One of the most counterintuitive findings involves a tactic that has exploded across B2B content marketing. Many companies now publish self-promotional "best of" lists, ranking themselves first and expecting AI to follow suit. Ahrefs researcher Mateusz Makosiewicz ran a controlled experiment publishing 34 such lists across five domains and tracking 9,886 AI answers. The result was sobering.

AI platforms used the articles as source material but did not recommend the brand behind the list. In one striking example, Makosiewicz published a list promoting Ahrefs' own conference. In 43% of AI-generated answers, the tool recommended a competitor's event instead. Glen Allsopp's separate analysis of 750 ChatGPT prompts found that "best X" list pages made up 43.8% of all cited page types — confirming AI's appetite for this format while exposing the gap between citation and recommendation.

The practical implication is significant: publishing a "best of" list may actively funnel AI-driven attention toward your competitors while your brand underwrites the content that makes it happen. Marketers running this tactic at scale should audit which brands are actually being recommended within their own published lists before continuing to invest in the format.

llms.txt Files Are Being Written for an Audience That Does Not Exist

Following Google's May 2026 guidance suggesting marketers audit their llms.txt files, the Ahrefs team examined server logs and bot traffic across 137,000 sites. They found that 28% of those sites had published an llms.txt file. Of those files, a staggering 97% received zero fetches. Among the small fraction that were accessed, 77% of visitors were not AI bots at all — they were SEO audit tools and platforms built specifically to study llms.txt adoption.

The finding exposes what Ahrefs describes as a "tail-eating snake effect" where SEOs produce files that only get read by the tools designed to track them.

This is a pattern worth taking seriously. The llms.txt rush mirrors earlier waves of structured data optimisation and meta keyword stuffing — tactics that generated industry-wide activity without proportionate returns. Before allocating resource to llms.txt implementation, teams should verify whether any AI crawler is actually accessing their domain at all.

Schema Markup Is Not an AI Citation Shortcut

A trend of "GEO experts" crediting quick AI visibility gains to schema optimisation prompted Ahrefs to run a controlled test tracking 1,885 pages that added JSON-LD schema against 4,000 control pages over 30 days. The result showed no meaningful citation uplift on Google AI Mode or ChatGPT. AI Overviews showed a small decline of 4.6%, though both treated and control pages were already trending downward before the experiment began.

Schema markup remains valuable for traditional search presentation — rich results, knowledge panels, and structured snippets — but the data does not support the claim that it accelerates AI citation. Marketers being sold "GEO schema packages" should ask to see controlled evidence before purchasing.

Classic SEO Still Does the Heavy Lifting

Traditional Search Rankings Remain the Gateway to AI Citations

Analysis of 1.4 million ChatGPT prompts found that 88.46% of all citations came from the general search index. Ranking well in classic search is still the primary driver of AI citation — a finding that directly contradicts the narrative that traditional SEO has been rendered obsolete by the rise of AI-powered search.

However, ranking on page one is not a guarantee of AI visibility. Analysis of 863,000 SERPs and 4 million citations found that only approximately 38% of URLs cited in AI Overviews also ranked in the top 10 for the same query — down from roughly 76% a year earlier. Google's increasing use of "query fan-out" means AI pulls citations from related searches rather than the user's original prompt, making topical breadth more valuable than single-keyword dominance.

The shift toward query fan-out has meaningful consequences for content planning. A site that ranks strongly for one primary term but has thin coverage of adjacent topics is increasingly exposed to AI citation gaps — even if its core page performs well in traditional results.

AI Answers Are Stable in Substance Even When the Words Change

Tracking 43,000 keywords over a month, with each AI Overview checked more than 16 times, revealed significant surface-level volatility. Wording changed 70% of the time, brands shifted 46%, and cited sources swapped 45.5%. Yet when semantic meaning was measured rather than exact wording, AI Overviews scored 0.95 out of 1.0 for consistency.

Google has settled on what the answer should be — it simply keeps rewording and re-sourcing it.

This distinction matters for how teams interpret AI monitoring data. Fluctuations in cited URLs or brand mentions within AI Overviews are not necessarily evidence that rankings have shifted in a meaningful sense. The underlying answer is largely fixed; the surface expression is not. Monitoring tools that report on exact-match citation changes may be generating noise rather than signal.

The Traffic and Click Reality Check

Google Still Sends 190 Times More Traffic Than ChatGPT

Studying 76,000 sites connected to Ahrefs Web Analytics, researcher Patrick Stox found that Google accounts for nearly 40% of site traffic while ChatGPT drives just 0.21%. ChatGPT also captures only approximately 12% of Google's search volume when non-search uses like coding assistance and translation are excluded.

Google is not being displaced quietly, and the data makes that plain. For brands reallocating budget away from search optimisation toward AI-platform visibility, this disparity demands scrutiny. The channel generating 190 times more traffic still warrants proportionate investment.

AI Overviews Are Cutting Click-Through Rates — Not Boosting Them

In May 2024, Google Search head Liz Reid claimed that links inside AI Overviews receive more clicks than traditional listings for the same query. Ahrefs researcher Ryan Law tested that claim and found the opposite. His 2025 study found AI Overviews reduced clicks to top-ranking content by 34.5%. By 2026, a repeat study found that the presence of an AI Overview cut position-one CTR by 58%.

That figure was corroborated by multiple independent sources:

Seer Interactive recorded a decline of more than 65.2%
Kevin Indig reported drops of more than 50%
Authoritas reported a decline of 47.5%

The convergence of evidence across independent researchers makes this one of the most robustly supported findings in the study. Brands that measure success by rankings alone — without tracking actual click volume — risk operating on a flattering but misleading picture of their search performance.

Understanding how search experience optimisation is changing the relationship between visibility and traffic is increasingly essential for any team relying on organic search as a growth channel.

Brand Mentions Outperform Backlinks for AI Visibility

Analysis of 75,000 brands, correlating multiple metrics against brand mentions in ChatGPT, AI Mode, and AI Overviews, produced a clear hierarchy of influence:

Domain Rating: weak correlation — 0.266 to 0.326
Number of backlinks: barely registered — 0.191 to 0.244
Branded web mentions: strong correlation — 0.656 to 0.709
YouTube mentions: strongest correlation — approximately 0.737

Links built authority for Google. For AI, what matters is whether people are genuinely talking about a brand across the web and on video platforms.

This finding reframes where brand investment should flow. A company with a modest backlink profile but strong community presence, active press coverage, and consistent YouTube visibility may outperform a heavily link-built competitor when it comes to AI citation. The authority signals that shaped a decade of SEO strategy are not the same signals shaping AI recommendations.

For a broader view of how these shifts connect to wider changes in search behaviour, the Search Engine Journal's coverage of AI and search evolution provides regularly updated research and practitioner perspectives worth tracking alongside studies like this one.

What This Research Means in Practice

Taken together, these findings point toward a strategy built on fundamentals rather than tactics. The brands that surface consistently in AI-generated answers are not winning through technical workarounds — they are winning because they are already well-represented across the web in ways that predate AI search entirely.

Marketers can apply this research in three practical ways:

Audit where brand mentions are actually appearing — across YouTube, forums, and independent press — rather than counting backlinks as a proxy for AI visibility.
Build topical content clusters that cover related queries comprehensively, rather than optimising narrowly for a single keyword, to account for AI query fan-out behaviour.
Measure true click loss from AI Overviews directly using analytics, rather than accepting platform claims at face value — the gap between what Google reports and what independent research finds is substantial and consequential.

In AI search, the brands that win are the ones people are already talking about. The data does not reward tactical shortcuts. It rewards genuine presence, topical authority, and the kind of sustained visibility that no single optimisation tweak can manufacture.