Google rankings improve the odds of AI citations, but they are not a universal gate. See what a 4,953-page experiment found across Google Search Console, Bing AI Performance, crawler logs, referrals, and commercial interest.
No. In a 4,953-page experiment on a brand-new domain, Bing reported 121,900 AI citations in three months while Google Search Console showed an average position of 52.5 and 63 clicks. Rankings still raise the odds: AirOps found pages ranking first in Google were cited by ChatGPT 3.5 times more often than pages outside the top 20, while Google’s AI features require an indexed, snippet-eligible page.
If your pages barely appear in Google and you're wondering whether to wait for rankings before expecting ChatGPT or Copilot to cite you, you don't have to wait. You also can't skip SEO.
In July 2026, I built a new directory of marketing tools to test that question on a site with no rankings to lean on. It holds alternatives, comparisons, category pages, tool profiles, and team-specific recommendations, mostly for bottom-of-funnel questions. I'm keeping its name private while the test runs, because naming it could draw the searches, links, and submissions that would change what I'm measuring.
Results at a glance
Signal
Observed result
Pages published
4,953
Initial Google visibility
Nearly 11,000 impressions per day within days of launch
Later Google visibility
918 impressions on July 20, then below 100 on many days
Recorded GSC totals
107,000 impressions, 63 clicks, 0.1% CTR, and average position 52.5
Typical crawler activity
Roughly 4,000 to 5,000 requests per day two months after the decline
Peak captured traffic
26,000 total requests in one daily window
Bing AI Performance
121,900 citations in the selected three-month view
Commercial response
About five submission requests per week
These signals come from different tools and reporting windows, so they aren't one unified time series.
Why ranking, crawling and citation are different things
Av. António Augusto de Aguiar 122C 1050-046 Lisboa, Portugal
Minnesota
202 N Cedar Ave, Suite 1 Owatonna, MN 55060, USA
A Google ranking, a crawler visit, and an AI citation are three different events. Mix them up and you'll read a crawler spike as a win, or a weak ranking as a dead end.
A crawler request means a bot fetched a resource from your site.
Storage or retrieval means a system may have kept or surfaced that resource for a later task.
A brand mention means an answer named your business.
A grounded citation means an answer displayed an exact URL as a source.
A click means a person followed through to your site.
A crawler request isn't a citation. A citation isn't a visit. A visit isn't a customer.
If you want to measure exact citations, track the URLs shown in answers rather than relying on crawler traffic or brand mentions alone.
Bing defines a citation as a source displayed in an AI-generated answer. Its AI Performance report counts citations across Microsoft Copilot, AI-generated summaries in Bing, and selected partner integrations. It doesn't report where a citation appeared within an answer, and it doesn't describe ChatGPT, Gemini, Claude, or every other AI platform.
How the experiment was set up
I built the directory on a new domain with almost no external authority. That mattered because an established brand, existing backlinks, or years of indexed content could have hidden the relationship between visibility and discovery.
I used Claude to research and draft the pages, and the information was generally accurate. I checked a few dozen pages by hand, corrected the approach, then scaled the same structure across the site. The directory used familiar SEO foundations:
Clear heading structures and repeatable page templates.
Links connecting tool pages with alternatives, comparisons, and categories.
Content organized around products and buyer use cases.
Structured data and crawlable server-rendered pages.
Bottom-of-funnel topics rather than broad informational articles.
Crawlers could reach and read every page. What the site didn't have was authority, so it had nothing to rank on.
What happened in Google Search Console
Google gave the directory a short period of broad exposure, then sharply reduced the number of queries where it appeared. The site launched on July 8, 2026, and Search Console showed close to 11,000 impressions per day within days.
On July 20, daily impressions fell to 918. They dropped below 100 on many later days. Across the period shown, the site recorded 107,000 impressions, 63 clicks, a 0.1% click-through rate, and an average position of 52.5. That is exposure beyond page five, not valuable search visibility.
Distinct queries fell from 6,212 to 136 while average position stayed broadly similar. The evidence fits Google ending an initial new-site evaluation window better than it fits a confirmed penalty. The directory remained indexed and crawlable, and the collected data didn't show deindexing or a confirmed manual action.
As Google's exposure kept shrinking, Bing's AI answers started citing the same pages.
What Bing's AI Performance report showed
Bing’s AI Performance report moved in the opposite direction. In the selected three-month view, it reported 121,900 citations and an average of 44 cited pages. On September 18, it showed 4,200 citations and 121 cited pages.
The chart begins close to zero in early August, rises during the middle of the month, briefly drops to zero in late August, and then settles into a band of roughly 3,000 to 4,000 citations into September. Those figures don't show that every citation produced a visit, or that every AI assistant cited the directory. They show growing citation activity in Bing’s AI ecosystem while Google showed very little useful visibility.
That result rules out a simple rule that says pages must first rank well in Google before AI systems can cite them.
What the crawler logs showed
Crawler activity remained substantial after Google’s visibility decline. Two months later, the site received roughly 4,000 to 5,000 crawler requests on an average day.
One captured daily window recorded 26,000 total requests, with 23,000 allowed. Google made 4,870 allowed requests, Microsoft made 3,040, OpenAI made 2,390, and Anthropic made 263.
Yes: founders and marketers sent about five requests a week to add tools, fix listings, or appear on alternatives pages. Some were automated, but others were clearly written by people who'd found a specific page and wanted to be on it. I can't prove which channel sent any one of them.
GA4 also began recording AI-assistant referrals. Those referrals eventually exceeded Google and Bing organic traffic combined, although the organic baseline had become extremely small. Raw session totals weren't useful because an unrelated scraper inflated overall property traffic.
Keep referral traffic separate from citations and actions. AI assistant referrals in GA4 can show whether people arrive, but citations and submissions answer different questions.
What other studies found about Google rankings and AI citations
Google rankings improve the chance of citation, but the effect changes by AI platform and query type.
AirOps, which sells AI search software, mapped ChatGPT citations against Google’s top 20 results across 43,233 queries: the original searches plus the related searches ChatGPT ran from them. It found that 55.8% of cited pages ranked in Google’s top 20 for at least one query. Pages ranking first in Google were cited by ChatGPT 3.5 times more often than pages outside the top 20. AirOps also found that 85% of sources ChatGPT retrieved were never cited. So ranking helps, but ChatGPT reading your page and ChatGPT citing it are two different steps, as AirOps’ study shows.
Ahrefs, looking at 3,311 classic head terms, found that only 10% of ChatGPT’s cited URLs matched Google’s top 10, far less than you would expect if ChatGPT simply echoed Google. Perplexity looked much closer to Google, with about 65% URL overlap with the top 10. Ahrefs’ analysis shows why you shouldn't assume that one AI platform behaves like another.
Rankability examined 1,645 AI citations and 916 traditional results across 31 topics. It reported that positions one to three had a 98.9% chance of citation, yet 55.2% of its top-10 AI citations didn't rank in the traditional top 10. Its findings show relationships in observed data, not proof that changing one page element will cause a citation. Rankability’s research also found stronger results for pages that introduced the topic early and covered it thoroughly.
Flozi’s much smaller study, covering four sites and 77 complete queries, cuts the other way on top positions. Its pages ranking one to three were cited in AI Overviews 31% of the time, against 53% for positions four to five, which Flozi thinks is likely a query-type effect. Pages ranking 11 to 20 fell to 12%. Flozi’s conclusion is that ranking well is necessary but not sufficient for an AI Overview citation.
For Google AI Overviews and AI Mode, Google says a page must be indexed and eligible to appear in Google Search with a snippet, with “no additional technical requirements.” Google sets no ranking threshold for eligibility. Eligibility only gets you considered, though: Flozi’s numbers show that rank still affects which eligible pages get picked.
These studies use different methods, query sets, and AI platforms, so don't add or compare their percentages directly. The experiment above is a different kind of evidence: one new domain followed over time, not a ranking-correlation study.
What the experiment does not prove
One domain can't settle this.
Its direct evidence comes from Bing Webmaster Tools AI Performance, Google Search Console, Cloudflare crawler logs, GA4 referral channels, and the directory inbox. It doesn't contain a controlled cohort of citation results across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews.
It does not prove that:
Google rankings never influence which sources AI systems choose.
A crawler request always resulted in storage, retrieval, or citation.
AI-written content caused the citations.
Another industry or an older domain would produce the same pattern.
A large-scale page rollout is safe or efficient for growth.
Submission requests came from one identifiable discovery channel.
The channels may influence each other. This experiment didn't establish that they are independent.
Why you should not repeat this on your main domain
Do not copy this publishing volume onto an established business website. I was willing to burn this domain. That was part of the test. Your main site isn't a test domain.
Publishing thousands of pages at once can weaken topical focus, create thin or repetitive pages, consume crawl resources, and expose a valuable domain to unnecessary risk. Accurate AI-assisted pages can still fail if they offer too little value or lack enough authority. SEO still matters, especially when you need dependable visibility from Google as well as AI search.
A safer route is to publish a small group of commercially important pages, measure what happens, and expand only when the evidence supports it.
What to do if you want AI to cite your pages
You don't have to finish your SEO roadmap before working on AI visibility. Measure how AI answers find and describe your brand now, and keep investing in search alongside it.
Make your pages crawlable and indexed. Google AI features need an indexed, snippet-eligible page. ChatGPT Search uses OAI-SearchBot to surface websites in search results, and OpenAI says sites that opt out won't appear in ChatGPT search answers.
Answer the main question early and cover it fully. Rankability found a 22.1% top-10 citation rate when pages introduced the topic within the first 100 words, compared with 8.7% otherwise. Thorough pages reached 27.6% versus 14.5%. That is correlation, not proof, but answering early and fully is a sound standard anyway.
Start with a small set of bottom-of-funnel pages. Alternatives, comparisons, and team-specific recommendations give you focused questions to answer. Build enough depth to help a buyer, rather than publishing pages simply to increase a count.
Measure each part of the chain separately. Track crawler access, grounded citations, referrals, and business actions. A rise in one doesn't automatically mean a rise in the others.
I am tracking questions such as:
Whether a website with little Google traffic can still appear in Perplexity.
Whether AI visibility can increase while Google impressions fall.
Whether rankings, links, or crawl access matter most for AI citations.
Whether a website needs Google rankings before ChatGPT cites it.
Each run records the AI platform, date, exact cited page, referral traffic, and any measurable action.
AI Peekaboo can track the prompts your buyers ask on the AI platforms you pick, up to seven: ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Microsoft Copilot, and Claude API. It keeps the answers and cited sources, with read-only GA4 and Search Console connections to compare AI visibility with search data. Start with a free AI visibility report to see whether AI answers name your brand today.
How we checked this
Last checked: 18 October 2026.
First-party evidence: a directory launched on July 8, 2026, with 4,953 AI-assisted pages across alternatives, comparisons, category pages, tool profiles, and team-specific recommendations.
Quality control: Claude researched and drafted the directory, then a few dozen pages were checked by hand and the approach was corrected before it scaled.
Measured sources: Google Search Console, Bing Webmaster Tools AI Performance, Cloudflare crawler logs, GA4 referral channels, and the directory inbox.
Interpretation: the site stayed indexed and crawlable, with no collected evidence of a confirmed manual penalty. Bing citation metrics and crawler activity remain separate signals.
Identity: the directory name, domain, URLs, analytics identifiers, and sender details remain withheld because the test is ongoing.
Limitation: this is one uncontrolled real-world domain, not a randomized experiment.
External source types re-read on 18 October 2026: third-party studies from AirOps, Ahrefs, Rankability, and Flozi; Google Search Central documentation; OpenAI crawler documentation; Bing Webmaster documentation; and Google US results for this keyword.
The study will be updated as its prompt cohort, citations, referral traffic, and submissions change.
Frequently asked questions
How do you rank in AI search?
You improve your odds by being crawlable, indexed where relevant, and useful enough to answer the question directly. Google AI Overviews and AI Mode need an indexed, snippet-eligible page. ChatGPT Search needs OAI-SearchBot access. Strong Google rankings help, but pages can still earn citations without reaching Google’s first page.
Are Google searches all AI now?
No. Google still shows ordinary search results alongside AI Overviews and AI Mode. Classic search eligibility still matters because Google's AI features can use supporting links from pages eligible to appear with a search snippet.
Does an AI crawler request mean a page was cited?
No. A crawler request means a bot fetched a resource. A citation means an answer displayed an exact page URL as a source. A visit is another event again, when someone clicks through. Measure those events separately because crawler traffic alone can't tell you whether your page appeared in an answer.
Did Google penalize the directory?
No confirmed penalty appeared in the collected data. The directory remained indexed and crawlable while distinct queries fell from 6,212 to 136 and average position stayed broadly similar. That pattern better fits the end of an initial new-site evaluation window than deindexing or a confirmed manual action.
Should you publish thousands of AI-generated pages to get cited?
No. This was a deliberate test on a domain I was willing to lose. On an established site, publishing thousands of pages can weaken topical focus, create thin or repetitive content, consume crawl resources, and put valuable Google visibility at risk. Start with a smaller set of important buyer questions instead.
Start with your own visibility.
See where your brand appears in AI answers and which competitors get mentioned.
Filipe Lins Duarte
