Do pages need author bylines to be cited by ChatGPT?

How many ChatGPT-cited pages show an author?

About one in three. Most cited pages in the audit named nobody, and complete credentials were rare.

Jacques and colleagues (opens in a new tab) drew 100 questions at random from HealthSearchQA, a set of 3,173 consumer health questions that Google Research compiled from real search suggestions. Each went into ChatGPT 5.2 Pro in a fresh account, with a request to include sources with links. The team then coded every cited page against an “Authority Signals Framework”, whose first question is “Who wrote it?”

So 43 of the 615 cited sources met the standard many content guidelines treat as essential. The audit is about health, where expertise matters more than almost anywhere, which makes the result more striking.

Which cited sources skip bylines, and why does it work for them?

Large institutions, whose name does the work a byline would do for a smaller publisher.

Citations were heavily concentrated. Ten organizations accounted for 52.8% of all health citations, led by Wikipedia at 10.7%, Mayo Clinic at 9.9% and Cleveland Clinic at 9.8%. Wikipedia is written by many contributors rather than one named author. Hospitals and government agencies carry the authority of their own names. The same pattern of a few sites winning most citations shows up in AI Mode and Perplexity.

Overall, 75.7% of citations came from sources with inherent institutional authority. The authors conclude that these sources rely on their standing rather than on explicit signals. Authorship, editorial process and technical optimization acted as secondary signals behind “who published it”.

Our own research points the same way outside health. In our Wikipedia and schema study, the strongest predictor of a brand recommendation was independent coverage. That means how many independent sites named the brand in the pages an assistant cited. Each tenfold increase went with 4.7 times the odds of a recommendation. Reputation built across the web, not a line on the page, carried the weight.

If not bylines, what did cited pages show instead?

Fewer quality markers than most guidelines assume, and technical polish more often than editorial ones.

Schema markup, the machine-readable labels that describe a page to software, was the most common signal. Editorial signals such as references and a stated medical review were present on a minority of pages.

Commercial health publishers were the exception. Lacking an institutional name, they stated a medical review on 71.1% of their cited pages. That suggests a reviewer statement may matter more to a publisher without a famous brand, though the study does not test it.

The audit also ran a sample of cited pages through an AI-text detector. Of 305 pages it could classify, 22.6% were judged AI-derived. A named human author was plainly not a condition of being cited.

Do bylines help in Google’s AI Overviews?

Not measurably, once pages competing on the same search are compared.

Our study of 3,096 pages ranking in Google’s top 10 compared the pages AI Overviews cited with those beside them that were not cited. A first, simpler comparison suggested pages with an author signal were cited 4.8 points more often. Comparing pages only with others on the same search, with all features entered together, turned that into −2.1 points, which is indistinguishable from no effect.

Ranking position, by contrast, mattered a great deal. Pages in positions 1 to 3 were cited about twice as often as those at 7 to 10. This is Google’s AI, not ChatGPT, but it is the only published comparison of cited and uncited pages on authorship we know of. We cover the other on-page signals linked to AI citations separately.

So are author bylines pointless?

No. They still serve readers and supporting evidence; there is simply no proof that they drive AI citation.

A byline with real credentials helps a human reader judge a page, and buyers often check sources after an AI answer. The audit authors, citing a 2006 national survey, note that 88% of US adults lack proficient health literacy, so many readers struggle to judge credibility at all.

Evidence inside the page appears to count for more than a name above it. In a 252,000-trial controlled test by Vishwakarma and colleagues (opens in a new tab) at Sprinklr, a software vendor, AI models preferred pages that backed claims with evidence such as tests or certifications. Publisher names were removed in that test, so the content alone made the difference.

What should you do about it?

Keep honest bylines where you have real experts, but do not treat them as the lever for AI visibility.

1. Do not hold back publishing because a page lacks a credentialed author; cited pages mostly had none.
2. Where a qualified person wrote or reviewed a page, name them with credentials and a review date. It helps readers and costs little.
3. Put the evidence in the text: references, figures, test results and clear sourcing for claims.
4. Build the publisher’s standing through independent coverage and search ranking, which the research ties more closely to citation.
5. If you are a smaller publisher in an expert field, show your vetting process openly, as cited commercial health sites do.

If you want help deciding where those efforts pay off for your category, see our generative engine optimization service.

What does the research not tell us yet?

No study has added bylines to real pages and measured the effect on ChatGPT citations.

• The ChatGPT audit describes cited pages only; it has no group of uncited pages to compare with.
• It covers health questions on one ChatGPT model, collected on January 11, 2026, with a prompt that asked for sources.
• Attribution was coded largely by an automated pipeline, with some fields checked by hand.
• Our AI Overview comparison covers Google’s AI, one day and US searches, and detects only author markup and on-page signals.
• Whether named experts help more in finance, law or business software is untested.

Frequently asked questions

Does ChatGPT check author credentials before citing a page?

There is no evidence that it requires them. In an audit of 615 ChatGPT-cited health sources, only 7.0% showed an author name with credentials, and 64.7% named no author.

Is E-E-A-T important for AI search?

The studies here did not test Google’s E-E-A-T guidelines directly. They do show that institutional reputation, independent coverage and evidence on the page went with citation far more clearly than bylines did.

Should I add author bios to my blog for AI visibility?

Add them if they are real and useful to readers, not as an AI tactic. Our study of AI Overview citations found no measurable edge for author signals once pages on the same search were compared.

What matters more than a byline for AI citations?

Who publishes the page and how well it ranks. In the ChatGPT health audit, 75.7% of citations went to institutional sources, and in our Google study, ranking position outweighed every on-page feature.

Sources

• Jacques, Datuowei, Jones, Basch, Vanderpool, Udeozo and Chapa (2026), Authority Signals in AI Cited Health Sources: A Framework for Evaluating Source Credibility in ChatGPT Responses (opens in a new tab), arXiv:2601.17109.
• Vishwakarma, Kumar and Jamidar (2026), What Gets Cited: Competitive GEO in AI Answer Engines (opens in a new tab), arXiv:2605.25517.
• Underneath (2026), What pages cited by AI Overviews have in common: 3,096 pages
• Underneath (2026), Do Wikipedia and schema make AI assistants recommend a brand?