News and Publisher Visibility in Meta AI

Meta AI's relationship with news is shaped by a company that has been deliberately retreating from the news business: Meta ended news availability in Canada in August 2023 in response to the Online News Act, and shut down the Facebook News tab in the US and Australia in April 2024. For publishers this means Meta AI surfaces news through general web retrieval — a Bing-backed index with freshness weighting — rather than through licensing deals or a curated news product, and visibility follows the same mechanics as any other content: rank, extractability, and crawler access.

How does Meta AI handle news queries?

Current-events prompts trigger grounded search, where recency and clarity decide which articles get quoted and linked. Freshness weighting is real but shallow: the assistant prefers the most recent authoritative account, so being first with a clean, extractable summary beats being definitive six hours later. For evergreen explainers ("what is the debt ceiling"), retrieval behaves like normal informational search, and well-structured explainer pages from news domains perform strongly — often better than their article archives, because explainers are written as self-contained answers.

The publisher decision stack

Which technical signals matter most?

Three things carry disproportionate weight. First, NewsArticle schema from schema.org with accurate timestamps — retrieval systems use datePublished and dateModified to arbitrate freshness, and publishers who backdate or omit them lose recency battles they should win. Second, crawler access: audit your robots.txt and CDN rules for meta-externalagent (Meta's AI crawler, documented at developers.facebook.com) and for Bingbot, since the Bing index is the retrieval backbone. Blocking Bingbot to protest AI answers removes you from Meta AI, ChatGPT search, and Copilot simultaneously — a decision to make deliberately, not by inherited config. Third, server-rendered article HTML: text injected client-side may never reach the fetcher.

How should publishers think about referral traffic?

Answer engines compress clicks; some publishers see AI-driven answers satisfy the query with no visit. The defensible position is to become the cited source rather than the summarized-and-forgotten one. Original reporting, exclusive data, and named-expert analysis are the content classes assistants must attribute because no secondary source can replace them — commodity rewrites of wire stories are precisely what gets synthesized away. Structure originals with quotable, fact-dense paragraphs (the 40–80 word passage unit) so the citation carries your name even when the click doesn't come.

Measure the exposure directly: run a weekly prompt set covering your beats ("what happened with X this week", "explain the Y ruling") and log which outlets Meta AI cites. Citation tracking across those prompts shows your attribution share versus competing outlets and whether policy changes — a paywall adjustment, a schema fix, a crawler decision — moved it. Treat that share the way you treat search rankings: a channel metric to manage, reviewed monthly against the mention-tracking baseline.