ChatGPT flags 'inauthentic mentions' on self-ranked listicles

On 5 September 2026, Metehan Yeşilyurt posted a screenshot of ChatGPT appending a note about "inauthentic mentions" to a list-format answer. He says he lost the citation despite ranking himself first in the list.

Yeşilyurt reports the pattern is rare so far. He attributes it to A/B testing of GPT-5.6 and an Astra 6 variant, alongside classifier updates visible in Statsig. The evidence is one screenshot from one operator.

Why this connects to Rudzki's August measurement

Three weeks earlier, Tomek Rudzki measured listicle citations roughly halving after GPT-5.6 shipped. That was a share-of-voice reading across a large sample. Yeşilyurt's screenshot is a single case, and it shows a mechanism that could produce that drop.

A classifier appears to be reading ranked lists, deciding some entries look self-promotional, and suppressing the citation while the text still renders. We think Rudzki and Yeşilyurt are seeing the same shift from two angles. The aggregate drop and the single enforcement event line up.

The seeded-at-launch problem

Neither observation on its own is proof. Together they suggest the model is derating a subset of listicles where the author is also the top-ranked entity. Any list seeded at launch with the publisher in first place fits that shape exactly.

The uncomfortable part for practitioners is that the classifier appears to look at structure rather than tone. A self-ranked listicle rewritten with humbler adjectives keeps the same shape. The safer move is to remove yourself from the ranked positions, or to publish rankings you did not author yourself.

Verify before you rewrite

Pull the last 90 days of ChatGPT citations for any listicle you own where you appear in the ranked positions. Compare citation rate before and after roughly 15 August, when Rudzki's drop began. A local drop tells you the classifier is touching your pages, and the fix is structural rather than cosmetic.