Which content types does ChatGPT cite most? A ranked breakdown

Product pages are the most cited content type in ChatGPT, at 22.4% of classified citations across all prompts. Landing pages come second at 16.8%, which means 39% of what ChatGPT cites describes what a company sells. Listicles follow at 13.6%, how-to guides at 8.7%, documentation at 7.5%, general articles at 6.1%, comparisons at 5.9%, news at 5.7%, FAQs at 3.3% and reviews at 1.7%.

Those figures come from Promptwatch's ChatGPT citation share by content type report, published October 6, 2026, covering September 28 to October 4, 2026. It classifies the pages ChatGPT cited and splits them by the intent of the prompt. The split is where the useful part is.

The ranking changes with intent

The overall order is a blend of three very different kinds of question.

Commercial prompts ("best X for Y" style questions where someone is choosing) are the most concentrated. Product pages take 26.0%, landing pages 20.6%, listicles 15.0% and comparisons 7.9%. Those four formats together account for almost 70% of commercial citations.

Informational prompts spread out more. Product pages still lead at 19.1%, with landing pages at 13.9%, listicles at 13.6%, how-to guides at 10.7%, documentation at 8.5% and general articles at 7.7%.

Branded prompts, where the user names a company, put almost half the weight on the company's own pages: product pages at 23.3% and landing pages at 22.8%. Branded prompts also lean on formats that barely register elsewhere. Reviews reach 5.6%, three times their overall share. Social posts reach 3.0%, almost four times their overall share. FAQs get 4.8% and press releases 2.1%.

One line stands out across all of it: product pages are the most cited type for every one of the three intents.

Taken as a whole, editorial formats (listicles, how-tos, articles, comparisons, news, reviews and other editorial) add up to 43%. Listicles, how-tos and comparisons alone make 28%.

What this does not tell you

The report is a single week for one engine, built from one vendor's classified citations. It does not tell you that a product page is cited because it is a product page. A product page might be cited because it carries the specs, the price or the availability that the question needs, and the format is just where that information tends to live.

It also does not cover other engines. The same classification on Google AI Overviews or Perplexity could look different, and nobody should assume the order carries across markets or languages. And a share of citations is not a share of clicks.

The label "product page" also hides variation. A thin page and a detailed page are in the same bucket. The report ranks formats, not quality.

What to do with it, no tool required

Treat the ranking as a prompt for an audit rather than a template. Here is how we would run that audit on a normal site.

Check the pages that describe what you sell. If product and landing pages draw this much citation share, they are doing more work than most teams assume. Look at each one as if you were the model answering a buyer's question: are the specs, the price, who it is for and any limits stated plainly, in text rather than inside an image or a script?
Write one comparison and one listicle for the questions your buyers ask. Commercial prompts give comparisons 7.9% and listicles 15.0%. If a competitor owns the "best of" page for your category, you are not in the answer, however good your product page is. Our post on whether listicles still work for AI search covers the format in more detail.
Keep one how-to guide per core use case. Informational prompts give how-tos 10.7%, well above the 8.7% overall.
Fill the branded gaps. If someone asks about you by name, ChatGPT leans on your product and landing pages, and also on reviews and social posts. Make sure your review profiles are complete and that you are not leaving the question to a single outdated third-party page.
Do not cut documentation. At 8.5% of informational citations it is a quiet performer.

For a longer view of how the mix has moved, our earlier look at product pages overtaking listicles shows the July shift, and the report here is a later, different window.

How to find your own mix

A content-type report describes everyone's citations pooled together. What you need is your own. When we rank tools for this job, classification by content type is the first thing we look for, and Promptwatch is the one we put first.

Its Citation Trends view classifies every citation by content type and source type and charts the citation rate over time, which lets you see whether it is your product pages, your guides or a third party's listicle that earns the citations on your prompts. Per page it shows when the page was first cited, how many days it took to ramp up, the peak rate and the decay half-life. Content gap analysis scores your coverage against the prompts you track, so you can see where a missing format is the problem. Prompt types (organic, brand and competitor comparison) and intent labels are built in, which mirrors the report's split closely enough to compare your pattern with the pooled one.

Once you know which format you are missing, the Content Agent (Business plan, $579 a month, no trial) plans, writes and publishes against gaps on a schedule, with a review inbox if you want a human check. Outside Business, the product also offers on-demand content creation from a brief and Content Refresh for existing pages, but the fact sheet we use does not tie those to a specific plan, so confirm that on the pricing page. Crawler logs, which show whether ChatGPT's bots reach those pages at all, start on Professional at $245 a month.

The takeaway

The ranking is product pages, landing pages, listicles, how-to guides, documentation, then the rest. It does not say product pages are enough, and it does not say your category matches the average. It does say that the pages describing what you sell carry unusual weight, and that comparisons and listicles matter more once the buyer is choosing. Audit those first, then measure what ChatGPT cites for your own prompts.