GEO for News & Media Sites: How Time-Sensitive Content Gets Cited by AI in Real Time
News GEO and evergreen GEO run on two completely different logics: evergreen content wins by being “remembered” by AI over the long term, while news wins by being “fetched” by AI in the moment. When ChatGPT, Perplexity, or Google’s AI Overview answers a question about an unfolding event, they retrieve the freshest pages live—they do not reach into the model’s stale memory. So whether you make it into a live answer comes down to three machine-readable signals: clear timestamps, event entity markup, and stable URLs. Miss any one of the three and AI may skip your exclusive scoop and cite a competitor who got all three right.
Why news GEO can’t follow the evergreen playbook
Because for time-sensitive queries, AI takes the “retrieval” path, not the “memory” path. When a user asks “what happened at this company today,” the model’s training data is long expired, so the engine dispatches crawlers live to grab the newest results, reranks them, and generates a sourced answer. This real-time pipeline cares about freshness and credibility signals—and the domain authority and internal-link networks that evergreen content optimizes for are of limited help here. A report published three hours ago with complete signals will beat a high-ranking old page that has no clear timestamp. So news sites aren’t fighting for “rank”—they’re fighting to be judged by the real-time pipeline as the most relevant, most credible source right now.
Requirement 1: clear, machine-readable timestamps
The first thing AI looks at to decide “is this the latest?” isn’t the date written in the body—it’s the structured time fields. In practice, get all of these right:
datePublishedanddateModifiedboth go into the NewsArticle schema, in ISO 8601 format (with timezone, e.g.2026-08-05T09:30:00+08:00).- The visible publish time must match the schema—don’t show “this morning” on the page while the schema says three days ago.
- Bump
dateModifiedon any major update, and note “Updated on…” in the body—in live queries, a fresherdateModifiedis often the deciding factor. - Don’t rely on relative time as your only signal: “2 hours ago” is friendly for humans, but machines want an absolute timestamp.
In one line: the timestamp is news GEO’s first ticket in, and a vague or contradictory date is self-inflicted demotion.
Requirement 2: event and entity markup
Live answers are organized around a specific event or entity. AI first has to confirm “this article is about exactly the event, company, or person the user asked about” before it will cite you. To strengthen entity disambiguation:
- Fill in
aboutandmentionsin the NewsArticle schema, using entity URLs (Wikipedia, official site,sameAs) to name the subject of the article—not just plain text. - Name the core entity in full in the headline and opening paragraph, not just a pronoun or abbreviation, so the machine can bind it instantly.
- Add a background block for major events, spelling out “what this event is and which entities are involved,” so AI can attach your report to the correct event node.
- Mark up place and organization entities (
Place,Organization)—local or industry events especially depend on this.
Done well, entity markup effectively files your report precisely under the exact question AI is answering.
Requirement 3: stable, clean URLs
The moment the real-time pipeline fetches your page, the URL has to return the full body reliably and fast—this is the most overlooked yet most fatal link:
- Don’t change the URL once it’s live: changing the address without a 301 turns a source AI just noted into a dead link.
- Don’t lock the body behind JS: if the initial HTML doesn’t contain the text, most AI crawlers see an empty shell, and what they can’t fetch they won’t cite (technical details in AI Crawler Management).
- Don’t block real-time retrieval crawlers: if “retrieval” crawlers like OAI-SearchBot or PerplexityBot are blocked by robots.txt or your CDN, you exit the candidate pool for live answers entirely.
- Fast, no walls: real-time fetching prizes speed, and a paywall or anti-bot challenge makes the engine give up on your story.
News GEO vs evergreen GEO: the difference at a glance
| Dimension | News / time-sensitive | Evergreen |
|---|---|---|
| AI sourcing path | Real-time retrieval (fetch-and-use) | Long-term indexing + rerank |
| Most critical signals | Timestamp, entity, stable URL | Domain authority, internal links, depth |
| Importance of freshness | Extremely high (won in hours) | Moderate (quarterly updates suffice) |
| What the first paragraph says | Event facts: who, when, what happened | A self-contained definition or conclusion |
| Primary schema | NewsArticle (with date, about) | Article, FAQPage |
| How success is judged | Made it into the “right now” answer | Cited over the long term |
The two aren’t in conflict: a news site should also keep evergreen background/explainer pages to catch search, but for the time-sensitive report itself, the optimization focus must shift to the real-time-pipeline side. For general content-structure principles, see GEO Content Structure.
Pre-publish checklist (60 seconds)
- NewsArticle schema has
datePublished+dateModified(ISO 8601 with timezone). - Visible page time matches the schema, not just relative time.
about/mentionsname the core event and subject via entity URLs.- Headline and first paragraph name the core entity in full.
- Initial HTML already contains the full body (not JS-loaded).
- OAI-SearchBot, PerplexityBot, and Google-Extended are not accidentally blocked by robots.txt / CDN.
- The URL is permanent and won’t change after going live.
Frequently asked questions (FAQ)
Q1: Do news sites still need traditional SEO? Yes, but with a division of labor. Traditional SEO earns long-term rankings for your evergreen background pages; live reports need the real-time pipeline of timestamps, entities, and crawler access handled separately. The two are additive, not either-or.
Q2: If I just publish fast enough, will AI cite me? Not enough. Speed is necessary but not sufficient. Publish fast but lack a machine-readable timestamp, lock the body behind JS, or block retrieval crawlers, and AI still can’t fetch or cite you. All three requirements have to be in place together.
Q3: Do I have to use NewsArticle schema? Is Article okay? For time-sensitive content, use NewsArticle—its semantics explicitly tell the engine “this is time-sensitive news” and it supports date and event-entity fields. Article works but is semantically weaker; it expresses time and event signals less precisely than NewsArticle.
Q4: How do I know if my report made it into a live answer? While the event is hot, ask the relevant question in ChatGPT (search on), Perplexity, and Google’s AI Overview and check whether you’re in the source list. For a systematic approach, see How to Track AI Search Visibility.
Q5: Does paywalled news content still stand a chance of being cited? It’s heavily discounted. If the real-time crawler can’t read the body behind the wall, there’s nothing to cite. The compromise is to expose a sufficiently complete free summary block (with the core facts and schema) so AI can at least fetch a citable passage, then drive the click to the full story.
Winning at news GEO isn’t about writing longer—it’s about whether three machine-readable signals are in place: timestamp, event entity, and stable URL. GeoSeoToday breaks “getting cited by AI” into concrete, testable items—the main GEO entry point is at /en/geo/, and you can also read What Is GEO? A Complete Definition of Generative Engine Optimization and GEO Content Structure.