Does ChatGPT Ignore Your Google Rankings? What a 320-Query Citation Study Actually Shows
CiteLens, a small AI-visibility platform, published a benchmark on July 8 that deserves a closer read than the launch week gave it. The team ran 320 buyer queries through ChatGPT, Perplexity, Claude, and Google's AI Mode, then matched every cited source against the same query's Google and Bing organic top-10.
The engines did not behave like one market. Google's AI Mode pulled 93% of citations from Google's top-10. Perplexity pulled 89%, with correlation to Google ranking near 0.9. Claude cited Google's top-10 only 53% of the time and leaned on brand recognition (58% of picks had Wikipedia presence). ChatGPT was the outlier: 30% overlap with Google's top-10, correlation near zero, and fewer than 4% of its citations appeared in Bing's top-10 at all.
That split is the story: not "SEO is dead," but SEO is engine-specific again.
Does the CiteLens data prove SEO is irrelevant for AI search?
No. It proves one SEO playbook does not fit four engines. For Google AI Mode and Perplexity, ranking still gates citations almost as tightly as classic blue-link CTR. For ChatGPT, index position barely predicts inclusion. Claude sits between: less rank-driven, more entity- and recognition-driven.
Treat the sample honestly: 320 queries, three consumer sectors, Turkish market, not a global law. CiteLens also sells visibility dashboards, so skepticism is warranted. Directionally, though, the pattern matches what practitioners see when they test the same brand across assistants.
TLDR: SEO still matters for Google AI Mode and Perplexity; ChatGPT and Claude need separate authority tracks.
Why does this explain the iCrowd press-release playbook?
We were skeptical that wire syndication alone buys AI citations in our iCrowd AEO press release piece. This data clarifies the mechanism. Pickups that get indexed can help Google AI Mode and Perplexity, because both still lean on search-like retrieval. They do little for ChatGPT, which is not rewarding rank the same way.
If your GEO plan is "more syndicated releases," you may be optimizing for two of four engines while assuming the others follow. They will not.
TLDR: Syndication-heavy AEO helps rank-adjacent engines; it is a weak ChatGPT strategy.
How should you split GEO work across engines?
Run two parallel tracks instead of one checklist:
| Track | Engines | Primary levers |
|---|---|---|
| Index & crawl | Google AI Mode, Perplexity | Technical SEO, internal links, indexable HTML, E-E-A-T, fresh content |
| Entity & authority | ChatGPT, Claude | Consistent naming, third-party mentions, primary docs, verifiable facts |
Within the index track, static HTML on the edge still beats client-rendered shells for crawlers, a lesson we bake into every ChatAd Land build. Within the authority track, question-based pages with direct answers (the format this site uses) give models greppable facts to quote.
Before you commit a quarter of budget, baseline where your brand sits with an AI-search visibility tool like Searchable ↗ rather than trusting vendor benchmarks alone.
TLDR: Two tracks: index SEO for Google AI Mode/Perplexity, entity authority for ChatGPT/Claude.
What numbers should you log in your own tests?
Copy the CiteLens method at smaller scale:
- Pick 10–20 category questions (not thousands of long-tail variants).
- Record which URLs each engine cites.
- Compare to your Google top-10 for the same queries.
- Tag overlap rate by engine weekly.
If Google AI Mode overlap is 80%+ and ChatGPT is under 30%, your SEO wins are real, and insufficient for full GEO coverage. Pair citation logs with inference cost if you are running automated evals; see our LLM cost measurement guide so monitoring does not outrun budget.
TLDR: Run a 10–20 query overlap test weekly; expect different scores per engine.
What should you change this month?
Audit one money page. Check Google rank, AI Mode citation, Perplexity citation, and ChatGPT mention for the same query cluster. If rank is strong but ChatGPT never cites you, add entity proof (docs, consistent product names, third-party references), not another keyword paragraph.
If rank is weak but you want Perplexity pickup, fix crawlability and structured facts first. There is no single dial.
TLDR: Match the fix to the engine: rank work for Google/Perplexity, authority work for ChatGPT/Claude.