Where do AI engines get their news?
From three layers: licensed and crawled news publishers, wire services, and the open web including your own newsroom. Perplexity retrieves 100% in real time and cites at sentence level, so a breaking announcement can appear in answers within hours if the retrieval layer can fetch it. ChatGPT's OAI-SearchBot works against the Bing index with live fetches by ChatGPT-User, which makes Bing Webmaster Tools verification and IndexNow submission the fastest path to being current there. Google grounds AI Overviews and AI Mode against its main index. In all cases, if the announcement is not crawlable text, it does not exist.
Why does press visibility matter more in a zero-click world?
The Ahrefs February 2026 study of 300,000 keywords found position-1 CTR falls about 58% when an AI Overview appears and roughly 68% of sessions end without a click. For announcement queries — is X launching, did Y raise funding — the answer itself is the destination. Being cited as the source of the news is the equivalent of the old position 1, and it accrues to whoever publishes clear, corroborated, first-hand facts. This is the newsroom extension of digital PR for GEO.
What makes an announcement retrievable?
- Publish a first-party press page on your own domain the moment the news breaks, with a dated headline, a 40-80 word summary containing the key facts, and named spokespeople with titles.
- Lead with the news, not the adjectives. Engines extract subject, action, object, date. “Company X launches Y in market Z on date D” survives grounding; “revolutionary synergy” does not.
- Distribute through outlets engines already trust. Wire distribution plus two or three sector publications create the corroboration grounding requires.
- Keep every number consistent across your release, the coverage and your product pages — funding amounts, dates, user counts. Inconsistency is the top reason claims get dropped at grounding.
How should your newsroom be architected?
Give every announcement a permanent URL, an index listing that crawlers can traverse, and Article or NewsArticle structured data with datePublished and author. Never let old announcements 404; redirect them. Engines treat the newsroom as part of your entity record, so structure and freshness discipline from site architecture for AI and content freshness apply directly.
How does coverage turn into citations?
Each independent article about your news is both a retrieval candidate and a grounding vote. When three outlets state the same fact you stated, engines gain confidence and cite the clearest passage — ideally yours. This is why the press release and the press page must be the most extractable version of the story; see press releases in 2026 for the exact structure. Monitor which version engines choose by asking your announcement questions weekly during news windows and screenshotting the citations.
What pitfalls undermine press GEO?
First, announcements that live only on a wire service or a PDF: your own domain ends up absent from the answer about your own news. Second, embargoes handled carelessly — if the press page goes live after coverage does, engines anchor on second-hand framing. Publish simultaneously. Third, hype language without facts: models skip it, so the quotable sentence never exists. Fourth, neglecting corrections: if coverage misreports a figure and you never publish a correction, the wrong number can persist in answers for months, because grounding prefers the majority. Finally, teams forget that news decays — after the window closes, fold the durable facts into evergreen pages so the retrieval value persists.
Measure news performance inside your four-dimension standard: add announcement questions to the fixed 50-100 question set, run it monthly, and score Visibility, Citation Rate and Brand Perception with screenshot evidence. News windows are when Brand Perception moves fastest, for better or worse.