Does ChatGPT Cite Recipe Blogs?

Short answer: yes, and less often than they should for the sites that have not thought about it.

We have been testing our own recipes across AI assistants since the audit work started. Here is what we have observed, with the caveat that this area moves fast and nothing here is a stable ranking factor yet.

The 30 blog study

In September 2026 we ran 30 food blogs through ChatGPT, Gemini, Claude and Perplexity. Eight questions per blog, each written from the blog’s own latest posts, so every question was one that blog should have been the best answer to. 952 answers in total.

Six blogs never came up once, on any engine. The typical blog was cited on fewer than one in ten of its own questions. ChatGPT alone named 9 of the 30 blogs and cited a blog on 6 percent of answers. The names that showed up instead were Reddit, YouTube, Food Network, The Kitchn and Facebook.

If you want your own number, the free AI visibility score runs twenty questions like these on your site.

What we tested

We took recipes from our own archive, asked assistants the questions those recipes answer, and recorded whether we were cited, paraphrased without attribution, or absent entirely.

The pattern that emerged is less about content quality than about machine-readability.

What seems to correlate with being cited

Complete, valid structured data

The posts that got cited were disproportionately the ones with complete Recipe schema. That makes intuitive sense: a model retrieving ingredient lists and cook times finds them far more reliably when they are marked up than when they are buried in prose.

Answering the question early

Posts that state the answer in the first paragraph get quoted. Posts that open with 400 words of story before the useful content tend to get skipped in favour of a competitor who leads with the answer.

Clear heading structure

H2s phrased as questions appear to be retrieved more readily than decorative headings. “How long do you bake it” beats “The bake” for this purpose.

Not blocking the crawlers

Obvious, but worth stating: sites blocking AI user agents are not cited. That is a legitimate choice with real trade-offs, but it should be a decision rather than a default someone set for you.

What did not seem to matter

Word count. Publishing recency, within reason. Whether the post ranked in the top three of Google for the same query, which surprised us.

Whether the traffic is worth having

This is the honest part. Citations do not always produce clicks. A reader who gets the ingredient list from an assistant may never visit.

What citations do produce is presence at the moment of decision. And the work that makes you citable, complete schema and answering the question early, is the same work that improves your standing in conventional search. You are not choosing between them.

How to test your own

1. Pick five recipes you would expect to be authoritative on.
2. Ask three different assistants the question each recipe answers.
3. Record: cited by name, content used without attribution, or absent.
4. Compare the cited posts against the absent ones. Look at schema completeness and where the answer sits in the post.
5. Repeat monthly. The behaviour changes.

Frequently asked questions

Should food bloggers block AI crawlers?

It is a genuine trade-off and reasonable people disagree. Blocking protects content from being used without attribution. It also removes you from a surface where readers are increasingly asking questions.

Does being cited by AI help my Google rankings?

No direct link is established. The underlying work, structured data and clear answers, helps both.

What is llms.txt?

A proposed convention for telling AI systems how to use your content, similar in spirit to robots.txt. Support is not universal and it is early. Low cost to add.