Discussion starter: an entity SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Entity SEO
- Structured Data
AI-assisted articles: do you still name a human author, add an AI disclosure in markup, or both? Any search impact either way?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
The short answer is both, but they do different jobs and only one of them has anything to do with markup. Name a real human who reviewed the piece and will answer for its accuracy, and disclose AI involvement in visible text where a reader would reasonably wonder how the article was made. Machine-readable AI disclosure for article text has no documented search effect in either direction, because Google does not document a structured data property for it.
The search risk sits somewhere else: in publishing low-value content at scale, and in bylines that claim authorship nobody actually exercised. The level of disclosure should follow the level of AI involvement, and if you publish public-interest text in the EU there is now a legal reason to get the human review part right.
What Google has documented, and what it has not
Google’s guidance on AI-generated content (February 2023) still frames the position: Google says it rewards quality regardless of how content is produced, and that using automation mainly to manipulate rankings breaches its spam policies. On disclosure, it says AI or automation disclosures are useful where a reader might wonder how something was created, and suggests adding them where they would reasonably be expected. It also says that giving AI an author byline is probably not the best way to make AI involvement clear to readers.
Google’s helpful content guidance uses the same logic under its “Who, How, and Why” questions: is it self-evident who wrote the content, does the page carry a byline where one would be expected, and is it clear how the content was produced. None of that is described as a ranking signal. It describes what a trustworthy page looks like to a reader.
The policy side is where consequences live. The spam policies cover scaled content abuse, which targets many pages produced mainly to rank with little value for users, whatever the tool. Google’s quality rater guidelines were also updated in January 2025 so that raters can give the lowest rating to main content that is AI-generated or paraphrased with little effort, originality or added value. Raters do not change rankings directly, but the guidelines show what Google’s systems are meant to reward.
What Google has not documented: any ranking benefit from disclosing AI use, any penalty for disclosing it, or any structured data property for flagging AI involvement in article text.
A byline and a disclosure answer different questions
Think of a structural engineer’s sign-off. Software did most of the calculations, and the certificate may say which software was used. The engineer still signs, and the signature means “I checked this and I will answer for it”, not “I typed every number”. A byline on an AI-assisted article should mean the same: this person reviewed the claims, added what the model could not know, and is accountable if it is wrong. The disclosure is the method statement.
The analogy has a limit. Google cannot verify either one literally. It evaluates the page, the site and signals around the author (author pages, other work, mentions elsewhere). A byline helps because it connects the article to a person with a track record that readers, and possibly Google’s systems, can check. It does not help if the person is invented.
Watch for this:
Invented author personas with stock photos and made-up credentials are the biggest risk in this whole question, and it is a trust risk before it is a search risk. When publishers have been caught doing it, the damage came from readers and journalists, not from a ranking system. Google’s August 2026 update to its site reputation abuse guidance also lists authorship clarity among the factors it considers for third-party content, which is a reminder that who wrote something is part of how Google reads a section of a site.
Match the byline and disclosure to how AI was actually used
“AI-assisted” covers everything from grammar suggestions to fully automated pages. Treating those the same produces either pointless labels on human work or misleading bylines on machine work.
How AI was used Byline Visible disclosure Markup Research, outlining, editing help; a human wrote the piece The human writer An editorial policy page is usually enough Standard Article markup with the writer as author AI produced a draft; a subject expert rewrote it, fact-checked it and added first-hand material The expert, as author Short process note on the article Author is the expert; no special property needed AI produced a draft; a generalist did a light edit “Edited by” the editor, not “Written by” Clear process note, near the top or by the byline Do not mark the editor up as the sole author if that misstates the work Automated at scale (data summaries, listings, product copy) The organisation or desk, plus a named responsible editor Label as automated, and explain the data source Organization as author; the scaled content abuse question matters more than the label The third row is where most real sites sit, and it is the uncomfortable one. If a generalist lightly edited an AI draft on a topic they do not know, a “Written by” byline overstates what happened. The better fix is usually not the label. It is getting someone with actual expertise to review the claims, which improves the page and makes an honest byline possible.
The markup question: what exists and what Google reads
For articles, Google’s Article structured data documentation supports author details (name, a URL for the author’s page, and whether the author is a Person or an Organization). That markup helps Google understand who wrote the page. It is not an AI disclosure mechanism, and there is no documented Google property that says “this text was AI-assisted”.
The machine-readable disclosure standards that do exist are mostly about media. IPTC’s digital source type vocabulary (including a value for media created by a trained algorithm) is used in image metadata, and schema.org has adopted an enumeration based on it. Google uses IPTC metadata for labelling AI-generated images in some surfaces, and Merchant Center has its own AI-generated content policy for product images and product text. None of that carries over to how Google Search treats an article.
So a markup-level disclosure on article text is harmless and, as far as public documentation shows, invisible to ranking. If you add one for compliance or internal governance, do not do it by setting the author to an AI tool or to a fake Person. Keep the human (or organisation) as author and put the disclosure in visible text, where readers actually see it.
The legal angle for EU publishing
The EU AI Act adds a reason to take “human review” seriously. Article 50(4) requires deployers of AI systems that generate or manipulate text published to inform the public on matters of public interest to disclose that the text is AI-generated. The exemption covers AI-generated content that has undergone human review or editorial control, where a natural or legal person holds editorial responsibility for the publication. Article 50 applies from 2 August 2026. The 2026 Digital Omnibus changes delayed high-risk obligations, but legal commentary indicates most Article 50 transparency duties still applied from that date.
The practical point for SEO teams: a named editor who genuinely reviews the work is not just a trust signal for readers. For EU public-interest content it is the basis of the exemption, so the review needs to exist and be documented, not just implied by a byline. This is a summary, not legal advice; the scope of “matters of public interest” is exactly the kind of question to put to counsel.
Search impact: what the evidence supports
There is not enough evidence to say disclosure helps or hurts rankings, and no controlled public test isolating the effect exists that I can point to. The mechanism argument also points to “probably neither”: Google’s documented position is about content quality and spam patterns, and a sentence saying “AI helped draft this” changes neither.
Watch for correlation traps in anecdotes. Sites that disclose carefully also tend to be the ones investing in expert review, original material and editorial standards. If they perform better, the disclosure is a marker of the process, not the cause.
The effect on people is more plausible. Early academic research on news audiences, such as Toff and Simon’s work on AI labels, suggests that labelling content as AI-generated can lower perceived trust for some readers. That argues for careful wording (“Drafted with AI assistance, reviewed and fact-checked by [name], [role]”) rather than a bare “AI-generated” badge, and for placing the named reviewer next to the note so the reader sees accountability and method together.
How to test it on your own site
A ranking test is weak here: editorial articles are rarely templated enough for an SEO split test, and a before and after comparison gets swamped by core updates. The measurable effect is on readers, so test that.
- Hypothesis. A visible disclosure with a named reviewer does not reduce engagement or conversion compared with no per-article note.
- Setup. A user-facing A/B test (not served differently to crawlers) on a set of AI-assisted articles: variant A shows the process note by the byline, variant B relies on the editorial policy page only.
- Metrics. Scroll depth, newsletter or lead conversions, return visits, and complaints or corrections. Watch organic clicks in Search Console alongside, but do not expect the test to detect a ranking effect.
- Decision rule. If the note measurably hurts conversion and no legal requirement applies, move disclosure into a consistent, visible process section lower on the page rather than dropping it. If a legal requirement applies, disclose regardless and test only the wording.
A practical position
If this were my site, the byline would go to the person who would answer a reader’s complaint about accuracy, backed by a real author page. AI involvement would be disclosed in visible text scaled to how much AI did, with an editorial policy page explaining the review process. Article markup would name the human or the organisation as author, never an AI tool, and I would not add invented disclosure properties expecting a search effect. For EU public-interest content, I would keep a record of who reviewed what. Most of the effort should go into the review itself, because that, not the label, is what the legal exemption depends on and what Google’s quality guidance describes.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.