What do AI assistants cite for marketing analytics questions? We checked 90 answers

When we asked six AI engines 15 marketing analytics questions, 85% of the 412 pages they cited (352) were cited by only one answer, and ChatGPT cited none of the URLs in Google's top 10 for the same questions. In this run, the pages that got cited again and again were the ones that answered the exact question asked: a free checker for a "free tool" question, OpenAI's own docs for an OpenAI crawler question, a listicle for a "best tools" question.

We ran this on 5 October 2026 to get a baseline for our own work, and we're publishing all of it, including the part that isn't flattering: OneLence was named in 0 of the 90 answers.

How did we run the test?

We took 15 questions that buyers of marketing analytics, AI visibility, MCP and affiliate tools actually type or ask, and asked each one once on six engines: ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode. That gave 90 answers, and all 90 came back.

• When and where: 5 October 2026, US English, logged out, no personalization.
• How: with OneLence's AI visibility engine, which asks each question on each engine and records every page the answer cites.
• What we recorded: every page each answer cited, and Google's organic top 10 for the same question on the same day, so we could compare the two.
• How we matched URLs: by exact URL, ignoring query strings and "www".

These were the 15 questions:

How spread out were the citations?

Very. The 90 answers cited 412 distinct pages from 264 domains.

• 352 of the 412 pages (85%) were cited by only one answer.
• Only 14 pages were cited by three or more answers.
• The most cited pages appeared in 5 of the 90 answers. Nothing came close to "everywhere".

So for these questions there is no single page that "AI" trusts. Each engine, on each question, picked its own handful.

Does ranking in Google's top 10 get you cited?

In this run, it mattered a lot for Google AI Overviews and hardly at all for ChatGPT.

A few things stand out:

• AI Overviews was the only engine where about a third of cited URLs were also in Google's top 10. Half of its cited domains were.
• ChatGPT made 55 citations and none of them was a top-10 URL for the question. Only 5 of its cited domains appeared in the top 10.
• Perplexity cited the most, 10 pages per answer, but only 6% of them were top-10 URLs.
• Claude cited the least, and 5 of its 15 answers had no citation at all.

If your plan for AI answers is "rank in Google and the rest follows", this run says that holds mostly for AI Overviews.

Which sites got cited most?

Counted by the number of answers that cited the domain (trimmed to the top of the list):

Two details matter here.

Reddit and YouTube were cited only by Gemini, Google AI Overviews and Google AI Mode. ChatGPT, Claude and Perplexity didn't cite either in this run. So "get on Reddit" is advice about Google's surfaces, at least for these questions.

LinkedIn posts by individuals got cited on 5 of the 15 questions: tracking ChatGPT traffic, the OpenAI crawler comparison, MCP servers for marketing, ChatGPT apps for marketing analytics, and AI referral traffic in GA4. These were posts and one LinkedIn article by individual people, not company pages. All but one were cited by AI Overviews; the article was cited by Perplexity.

Which pages were cited by three or more answers?

These are the 14 pages that showed up in at least three answers:

What kind of page wins?

Three patterns show up in the tables above. They describe this run, not a law.

1. The page that matches the question exactly. The free-tool question was answered with free tools (crawlercheck.com, Siftly's crawler audit). The OpenAI crawler question was answered with OpenAI's own bot documentation. The Search Console and Shopify MCP questions were answered with GitHub repositories and MCP directories. The affiliate de-duplication question went to Shopify's and Awin's help docs. "How to track" questions went to how-to guides (Semrush, Surfer).

2. Listicles for "best X" questions. "Best AI visibility tools" was answered mostly from two "best tools" articles: one from Frase, a vendor in the category, and one from Zapier, a publisher. Each was cited by 5 answers. For "Peec AI alternatives", the cited listicles also came from vendors in the category (Geoptie, Writesonic).

3. The competitor's own home page for "alternatives" questions. For "Peec AI alternatives", Peec AI's home page was among the most cited pages, next to those vendor listicles. The Triple Whale question went a different way: its citations were spread across several vendors, two answers each.

What does this mean if you want to be cited?

Everything here is drawn from this one run, so treat it as a starting checklist, not a promise.

1. Write one page per question, in the question's words. The pages that showed up in several answers were the ones that answered one specific question directly: a checker for "free tool to check…", docs for "X vs Y", a guide for "how to track…".
2. Don't stop at Google's top 10. Outside AI Overviews, most cited URLs weren't in the top 10. For ChatGPT, none were.
3. Check the engine before you pick the channel. In this run Reddit and YouTube only appeared in Google's surfaces and Gemini. If ChatGPT or Perplexity matters more to your buyers, those threads and videos won't show up there.
4. Practitioner posts count. Individual LinkedIn posts were cited on 5 of 15 questions, mostly by AI Overviews. A clear, specific post from a person can sit next to vendor pages.
5. For "best X", the lists are the battleground. Vendor listicles and publisher roundups carried those questions. Being in the lists that already get cited matters as much as your own page.
6. Make sure assistants can read you at all. None of the above helps if your robots.txt or firewall keeps AI crawlers out.

What are the limits of this data?

• One run, one day, one country, logged out. Answers vary between runs, so treat every count here as a snapshot.
• 15 questions is a small set, chosen around the topics we work on. Other topics may cite very differently.
• No brand-mention counts. Our extraction of which brands each answer named wasn't complete enough to publish, so we only report citations, plus our own result: OneLence named in 0 of 90 answers and linked in 0.
• We'll repeat it quarterly with the same 15 questions and publish what changed.

Where does OneLence fit?

We build OneLence, which helps marketing teams find growth opportunities, know what to do next, and grow more efficiently without simply spending more. Our AI visibility page explains how we look at which pages AI crawlers read and the visitors assistants send. If you're choosing a tool for this, we compared 12 of them, with prices, in AI visibility tools compared.

FAQ

What do AI assistants cite for marketing analytics questions?

In our 5 October 2026 run, mostly pages that answered the exact question: free tools, official docs, how-to guides, and "best tools" listicles. 85% of the 412 cited pages were cited by only one of the 90 answers.

Does ranking in Google's top 10 get you cited by ChatGPT?

Not in this run. ChatGPT made 55 citations and none of them were in Google's organic top 10 for the same question. Google AI Overviews had the most overlap, with 36% of its cited URLs in the top 10.

Do AI assistants cite Reddit and YouTube?

In this run, only Gemini, Google AI Overviews and Google AI Mode did. Reddit was the most cited domain overall (13 answers) and YouTube second (12), all from those three.

Which engine cites the most sources?

Perplexity, with 10 citations per answer on average. Claude cited the fewest, 3.3 per answer, and gave no citation in 5 of its 15 answers.

How often will you repeat this?

Quarterly, with the same 15 questions, so changes are comparable.

Next step

Before anything else, check that AI crawlers can actually read your site. Our free robots.txt checker for AI crawlers shows which ones your robots.txt lets in.