← All posts
November 17, 2026

What Do AI Engines Actually Cite? 617,985 Citations, Broken Down by Content Type

Listicles dominate the pages AI engines cite, but that average hides the real story. Claude leans hardest on listicles while ChatGPT spreads citations across comparisons and docs. The format that gets cited depends on which engine your buyers use and what question they ask.

In short

  • Listicles dominate the pages AI engines cite overall, but that average hides the real story.
  • Claude leans hardest on listicles, while ChatGPT spreads citations across comparisons and docs, and Perplexity leans on forums.
  • The format that gets cited depends on which engine your buyers use and what question they ask.
  • Format is not a ranking hack: engines cite what they can verify, not what is structured as a list.

What the Otterly.AI study measured and where it stops

We looked at data from an Otterly.AI study published in October 2026 that tracked 617,985 citations across seven AI engines and five industries. Getfluence, which connects brands with publishers for paid placements, is a commercial partner of OtterlyAI, and OtterlyAI customers get 25% off their first campaign. The study covered US prompts only, and found most of those prompts asked for rankings or "best of" lists. Researchers classified the format for most citations, leaving some unclassified.

A snapshot like this is not a universal law, and engines update their retrieval behavior continuously.

Listicles dominate overall but swing hard by engine

The headline figure is striking: most cited pages are listicles when you combine all seven engines. That number has shown up in marketing advice already, usually as "write listicles." We compared the engine-level data and found the advice falls apart on inspection.

Claude shows the heaviest concentration, with the large majority of its citations going to listicles. ChatGPT spreads citations more widely: listicles still lead, but comparisons and docs take a meaningful share. Perplexity leans heavily on forums and discussion threads. Gemini balances across listicles and explainers. The remaining engines each show their own skew.

If your buyers use Claude, listicles matter more. If they use ChatGPT for product comparisons, a head-to-head post has better odds of being cited than a ranked list.

Choose

What kind of question are your buyers asking the AI engine?

Pick an answer to see what fits.

Write a listicle

Ranking prompts skewed citations heavily toward listicles. Example: "Best project management software 2026" calls for a listicle.

Write a how-to guide

Open-ended prompts lowered the listicle share and raised how-to guides.

Write a head-to-head comparison

"Vs" and comparison prompts made head-to-head comparison posts the most cited format. Example: "Asana vs Monday.com" calls for a comparison.

Write an explainer with technical depth

How-does-it-work prompts sent citations toward docs and explainers. Example: "How does automated task assignment work" calls for an explainer with technical depth.

Why format is not a ranking hack

The 66.6% figure has already been weaponized into content advice that treats listicles as a cheat code.

Engines cite what they can verify, not what is structured as a list. A thin listicle written mainly to sell, with claims that cannot be traced to a source, is unlikely to sustain citations as engine retrieval improves. The Otterly.AI study measured what was cited, not what was accurate.

The engine-level differences matter more than the headline average. Writing listicles because "most citations are listicles" ignores that ChatGPT cites them far less often, and that Perplexity barely touches them.

Steps

How ContentRails matches format to engine without rewrites

  1. Add your site and research topics

    ContentRails researches topics with real search numbers and writes listicles, comparisons, and explainers from your site's content. Project DNA profiles your audience and voice per site, so a comparison post for ChatGPT readers and an explainer for Perplexity readers both sound like you. [1]

  2. Pick the format when requesting a post

    Matching format to engine is a settings change: when requesting a post, you pick the type—listicle, comparison, explainer—not after drafting. Every figure is traced to a source, matching what engines verify. [1]

  3. Review the sourced draft

    Autopilot levels range from Off to Full autopilot on Pro, with human approval before anything publishes. [1]

  4. Approve or edit, then publish

    The Pro plan runs $100 a month for up to 5 sites and 20 posts; the Free plan covers 1 site and 3 posts. [1]

  5. Or test on your own site first

    The free trial writes one post in about 15 minutes with no card, reads your public pages only, and publishes nothing. Claim it within 24 hours to keep it. You can also try the free citation checker, which reads a published post and reports which figures have a source. [1]

Sources

Built on 1 source

Vendor 1
  1. [1]/for/saasVendor · contentrails.ai