News publishers lost about a third of their Google search traffic in 2025, and Google Discover now sends them more visitors than search. AI Overviews and AI Mode changed the economics of informational content, but they mostly stay away from breaking news. ChatGPT and Claude cite news sites, but send almost no clicks. This article collects what the data, Google's own documentation, the September 2025 Search Quality Rater Guidelines, and practitioners such as Barry Adams, John Shehata and Sophie Winkels say about news SEO in 2026, and ends with a checklist.
Every claim links to its source. Where a number comes from a vendor's own panel, the text says so. Where the evidence is thin, the text says that too.
Key takeaways
Search is shrinking, not dying. Chartbeat measured Google search referrals to publishers down 33% worldwide and 38% in the US in the year to November 2025, and down 40.2% in the year to July 2026. Barry Adams calls "Google Zero" a lie: Google is still the largest referrer by far, and no other channel is replacing it.
AI Overviews avoid hard news. Google says it aims not to show them for hard news topics. They appear on roughly 3% to 6% of news queries, concentrated in health, science and technology. Where they do appear, clicks fall by a third to a half.
Discover is the biggest Google channel for news, and the least stable. It delivers about two thirds of publishers' Google traffic. Google ran its first Discover-specific core update in February 2026 against clickbait and for topic-level expertise.
Breaking news is the one growing category. Define Media Group saw breaking-news traffic up 103% across Search and Discover since November 2024, while evergreen guides fell 40%.
AI-written journalism is a trust and quality risk, not a ranking trick. Google's rater guidelines rate scaled, low-effort AI content Lowest. Only 12% of readers are comfortable with fully AI-generated news.
Chatbots are a visibility channel, not a traffic channel. ChatGPT and Claude referrals are under 1% of pageviews. Allow their search crawlers, decide on their training crawlers, and measure citations separately.
What happened to publisher traffic: the numbers
The studies below use different panels and periods. Read the trend, not the decimals.
Finding
Source
Google search referrals to publishers down 33% globally and 38% in the US (Nov 2024 to Nov 2025, 2,500+ sites); Discover down 21% globally and 29% in the US
Google Search pageviews down 40.2% year on year (July 2025 to July 2026), Discover down 34.3%; search share of pageviews fell from 9% to 5%; AI referrals 0.01% of traffic
Users click a traditional result in 8% of visits with an AI summary versus 15% without; 1% click a link inside the summary (900 US adults, 68,879 searches, March 2025)
Breaking-news traffic up 103% across Search and Discover since November 2024 (Discover up 168%); evergreen, how-to and reviews down about 40% (64 US publishers)
Define Media Group, reported by Press Gazette, 18 March 2026
Traffic to the ten biggest US news sites down 32% in two years; organic search to the top 50 down 42% year on year; generative AI is 0.4% of referrals
Google disputes the framing. In August 2025 Liz Reid, Google's head of Search, wrote that "total organic click volume from Google Search to websites has been relatively stable year-over-year" and that click quality went up. The publisher-side data above comes from several independent panels, and all of them point the same way. The honest summary: aggregate clicks may be stable across the whole web, but clicks to news and information publishers are falling.
Barry Adams, the most widely read news SEO consultant, disagrees sharply. In "Google Zero is a Lie" (2 March 2026) he writes that the narrative "is entirely false, and extremely dangerous", that The Verge's own Google traffic was declining before AI Overviews launched, and that "there is no other channel rising to take Google's place. Choosing to abandon Google is a potentially catastrophic strategic error." His earlier piece "AI Survival Strategies for Publishers" (8 October 2025) frames it as "Peak Search": the total number of clicks Google sends will not grow, so "search is now a zero-sum game. When you get a click, a competitor isn't."
The SEO leads of the Daily Mail, The Telegraph and Bauer take a similar line in Press Gazette's "Search isn't dead, it's fragmenting" (5 February 2026). Carly Steven of the Mail notes AI Overviews appear on only about 12% of non-brand mobile queries in the UK and have plateaued since mid-2024. Her advice from May 2025 still stands: "Focusing on the brand has never, ever, ever been more important."
Our reading: this is a fact-versus-forecast dispute. The decline is a fact. Zero is a forecast, and the evidence for it is weak. Plan for less search traffic, not for none.
What Google itself has written
Google publishes a lot for publishers. The documents below are the ones that matter for news in 2026.
AI features and your website: "There are no additional requirements to appear in AI Overviews or AI Mode." AI Overviews "are only shown when our systems determine that it is additive to classic Search, and as such, often don't trigger." The only crawl control is the Googlebot robots.txt rule; nosnippet, data-nosnippet, max-snippet and noindex limit what is shown.
Optimizing your website for generative AI features on Google Search (May 2026, updated July 2026): "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." It explains query fan-out and grounding, says Google ignores llms.txt, says there is no need to "chunk" or rewrite content for AI, and warns that pages built for fan-out queries to manipulate rankings violate the scaled content abuse policy.
Search generative AI control in Search Console: since 31 August 2026 every site can exclude its content from AI Overviews, AI Mode and generative features in Discover. The control "isn't used as a ranking or inclusion signal affecting other parts of Search" and does not affect training; Google-Extended handles that. The UK Competition and Markets Authority forced this opt-out on 3 June 2026. Before that, the only way out of AI Overviews was leaving Search.
Generative AI performance reports in Search Console (June 2026, all sites since 31 August 2026): separate impression reports for Search and for Discover. No clicks, no queries.
Preferred Sources and "Highly Cited" (27 May 2026): users can mark preferred news sources, now also in AI Overviews and AI Mode. Google says people are "twice as likely to click through to a Preferred Source". A "Highly Cited" badge marks original reporting other stories cite. Google offers an embeddable button; by August 2026 users had picked more than 600,000 unique sources.
News topic authority (May 2023): Top Stories and the News tab weigh how notable a source is for a topic or location, how often its original reporting is cited by others, and its reputation.
News content across Google: "Publishers are automatically considered for 'Top stories' or the News tab of Search." No Publisher Center submission is needed; the Google News policies require clear dates, bylines, author and publisher information, and no disguised sponsored content.
Using generative AI content (updated 1 October 2026): "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." This now applies to titles, meta descriptions, structured data and alt text too.
Google's commitment to the information ecosystem (18 June 2026): Google pays more than 200 publications in its News AI pilot program and pilots payments to sites "whose content meaningfully contributes to the freshness and factuality of generative AI responses". Digiday reported publishers call the offers "peanuts".
AI Overviews, AI Mode and the Top Stories exception
NewzDash 2025 study (2 June 2025): 3.47% of trending US news queries trigger an AI Overview. By category: health 17.32%, technology 6.59%, science 4.44%, business 2.19%, sports 0.50%, world 0.31%, entertainment 0.20%. AI Overviews appear most often six to nine hours after a major news event.
Ahrefs (10 November 2025, 146 million SERPs): 6% of news queries trigger an AI Overview; "AIOs concentrate on evergreen queries, not current news."
Trisolute News Dashboard (December 2025): 11.80% of US SERPs with Top Stories also showed an AI Overview; in the UK 3.56%.
Semrush (December 2025): news and sports lag the average growth of AI Overviews, which "suggests Google is still wary of covering recent events with AI-generated content."
Barry Adams' explanation, from his WAN-IFRA deck "Google's AI Overviews: The New Search" (6 May 2025), is that a query that deserves freshness triggers a Top Stories box, and "QDF != AIO". This is a theory, and he labels it as one. Two things changed since. In July 2026 NewzDash found the Top Stories carousel embedded inside the AI Overview on 15.5% of US and 17.46% of UK trending news queries that show Top Stories, mostly in entertainment and world news. And AI Mode, which Google says passed one billion monthly users in May 2026, answers news queries with far fewer links. Adams observed AI Mode answering news questions with only Wikipedia links where the normal SERP showed seven news stories, and warned that "the shelf-life of news is reduced even further."
Practical consequence: the Top Stories box and Discover are where news SEO happens. Optimise article structure for grounding anyway, because explainers, health and technology stories do get summarised. Adams' recommendation is content segmentation: one topic per section, a strong subheading phrased as a question, and the complete answer in the first 150 words.
How Google Discover works
Discover is now the largest Google channel for most news publishers. Chartbeat data shows 68% of Google traffic to news sites comes from Discover (August 2025). NewzDash puts it at 67.5% for the fourth quarter of 2025. John Shehata's warning in Search Engine Journal: "Google Discover is very unreliable, but at the same time, it's addictive." It is a channel, not a strategy.
What Google says about Discover
Discover "shows people content related to their interests, based on their Web and App Activity" (Discover and your website, updated 9 March 2026). No query, so no keywords.
"Content is automatically eligible to appear in Discover if it is indexed by Google and meets Discover's content policies. No special tags or structured data are required." Eligibility "is not a guarantee of appearing."
Discover "makes use of many of the same signals and systems used by Search to determine what is helpful, people-first content."
Images: at least 1200 pixels wide, more than 300,000 pixels, 16:9. Allow large previews with <meta name="robots" content="max-image-preview:large">. Do not use the logo as the image.
Headlines: "capture the essence of the content", no clickbait, no "sensationalism tactics that manipulate appeal by catering to morbid curiosity, titillation, or outrage."
"Traffic from Discover is less predictable or dependable when compared to keyword-driven search visits." Google calls it supplemental.
The Discover content policies require clear dates and bylines, author and publisher information. Violations can trigger a Discover manual action in Search Console.
The February 2026 Discover core update was the first Discover-specific update Google announced. Goals: more locally relevant content from sites based in the user's country, less sensational content and clickbait, and "more in-depth, original, and timely content from websites with expertise in a given area". Expertise is judged "on a topic-by-topic basis": a local news site with a gardening section can rank for gardening, a movie site with one gardening article cannot.
Technical base: server response under 200 ms (under 600 ms is "good enough"), no client-side JavaScript for article content because "Google strongly prefers fresh articles" and rendering is delayed, semantic HTML, and minimal crawl waste. His example: a one-million-URL publisher crawl where only 8% of URLs were indexable.
Structured data: NewsArticle with exactly three images (16:9 at 1200x675, 4:3 at 1200x900, 1:1 at 1200x1200), the visible top image identical to the 16:9 one, lean snippets, Person and ProfilePage for authors, NewsMediaOrganization for the homepage.
Open Graph matters: "OG:title and OG:image are often used by Google as the headline and featured image in Discover. Optimise the og:title for Discover (and social) to maximise CTR."
Headlines: descriptive, accurate, with named entities, and a curiosity gap without misdirection. His example moves from "Hamilton: No decision yet about future" to "Life after F1: Ferrari's Lewis Hamilton undecided about his future".
Intro: a short summary with extra entities and topics, in semantic HTML, to widen the article's relevance.
Images: people, faces, dramatic scenes; a little text is fine. Test what drives CTR.
Engagement: "Discover is highly sensitive to interaction signals": bounce rate, time on site, pages per visit. Articles that get a lot of traffic from other channels in a short time "often see a boost in Discover."
Authority: site authority, topic authority and author authority. "How trusted are your journalists?"
The warning: "It works... until it doesn't." Clickbait and low-effort content accumulate low-quality signals that "eventually bear fruit in a Google core update", and "sites negatively impacted by a core update very rarely recover." His closing line: "Do not let Discover set your editorial strategy." His April 2026 essay "The Dangerous Seduction of Click-Chasing" is the long version.
Lily Ray's study of 83 US publisher sites over 16 months, presented at NESS 2025 and summarised by journalism.co.uk, found the highest Discover CTR for extreme weather, political predictions, celebrity financial drama and health scares, headlines of 90 to 105 characters, and images with human faces. Note the tension: the formats that win CTR are the ones the February 2026 update targets. The December 2025 core update already cut Discover traffic for some sites by 50% to nearly 100%, and Reach plc reported Discover referrals down 46% in the second half of 2025.
The Tactical Guide to Optimising for Google Discover (Belgrade SEO Conference, uploaded 18 September 2026, 46 slides). The full Discover playbook above, plus the new "Search profiles" follow feature (US first) and the NESS 2026 dates: 13 to 15 October 2026 at newsseo.io.
How to Optimise for Google Discover in the Age of AI (SEO for Paws charity stream, March 2026, 26 slides). Shorter version with the Hamilton headline example. Despite the title it has no slides on AI Overviews; the "AI" angle is Discover's shift to AI summaries and social content.
The Most Common Technical SEO Issues for News Publishers (News Reach, Dortmund, 24 October 2025, 46 slides). "Rendering takes time, and news doesn't have time. News article indexing is initially with raw HTML only." Pagination "is almost always done wrong"; indexable pagination builds topic authority. Core Web Vitals are "not a big direct ranking factor" but bad usability creates negative click signals. The final slides: "Technical SEO for LLMs... is technical SEO."
30 Years of SEO Bullshit (Nottingham Digital Summit, 25 September 2025). GEO listed as the latest fad. Closing advice: "Don't trust Google. Don't rely on tools. Test what works. Chase after people, not algorithms."
Managing Googlebot's Greed (Hive MCR, 15 May 2025). The priority crawl queue covers "news website homepages and key section pages"; robots.txt is the only crawl-management signal; "one piece of content = one URL".
Google's AI Overviews: The New Search (WAN-IFRA World News Media Congress, 6 May 2025, 52 slides). AI Overviews cite six links on average, mostly from page one. "AIO does not appear for news queries, with very few exceptions." Block Gemini training with Googlebot-Extended: "probably a good idea". Use nosnippet to avoid citation: "probably a bad idea". Conclusion: "AI Search will not replace classic search... It's still SEO."
Site Migrations for News Publishers (NESS 2024, uploaded March 2025). Top Stories and Google News inclusion are tied to the hostname. Changing domains means losing inclusion; "301 redirects are not the default solution, they are the last resort."
On LLMs his position is consistent and blunt. In October 2025: "Citations in LLM responses don't lead to clicks in any meaningful way, so the traffic opportunity there is non-existent." In an interview with Storycue (January 2026): "Journalism itself should never be automated."
Content with a short lifetime: live blogs, matches, breaking news
A live blog of a football match or an election night is relevant for hours, then nearly worthless. Breaking news is the only content category that grew on Google in 2025 and 2026, so this is where the effort pays. The guidance below combines Google's documentation with the advice of Barry Adams and the WTF is SEO? newsletter by Jessie Willms and Shelby Blackley.
Before and during the event
Publish early, on one URL. "URLs are sacred." Set a keyword-focused slug of about five words at publication and never change it. Google's priority crawler picks up news URLs "usually within a few minutes", so the article must be optimised at first publish: headline, intro, image and structured data in the raw HTML.
Say it is live. Put "live" or "updates" in the headline and in the first 70 characters. The headline at publication locks in the keywords you can compete for; change the headline as the story turns, because a changed headline is one of the triggers for a recrawl.
Update with substance. WTF is SEO? recommends updates of at least 100 to 150 words, no more than five to ten minutes apart at peak, and "paired updates": headline, intro, image, meta description and body together. Adams notes that Google keeps "live coverage articles (that are recognised as such by Google)" in the priority crawl queue for around 24 hours.
Use LiveBlogPosting markup, without counting on the badge. Google has no public documentation for LiveBlogPosting. Adams wrote in September 2024 that "there is no official Google documentation on LiveBlogPosting markup", that the Rich Results Test always shows an error, and that it "still seems to work fine", but "there's no guarantee that you get the Live badge". Minimum fields: coverageStartTime, coverageEndTime, author and liveBlogUpdate entries.
Pin the summary. Keep a short, current "what we know" block at the top. That block is what Top Stories, Discover and any AI summary will use.
Link it from everywhere. Homepage, section front, story hub, and the first mention in every related article. Topic pages signal "that your organization is committed to continuing coverage."
Keep dates honest. Google's publication date guidance: show one visible date, label it "Published" or "Last updated", keep it consistent with datePublished and dateModified, and never set a date to make a page look fresh. The helpful content questions ask it directly: "Are you changing the date of pages to make them seem fresh when the content has not substantially changed?"
Option A: convert the live page. Rewrite the same URL into a report with a new headline, a proper intro, and the timeline below. The URL keeps its links and signals.
Option B: new wrap-up, de-optimise the live page. Publish the match report or analysis on a new URL, link to it from the top of the live blog, and remove "live" from the old headline. WTF is SEO? uses this for long-running stories where the live format restarts daily.
Do not 301 every finished live blog. Adams: redirects are the last resort. Redirect only when the content is truly replaced. Recurring events (the annual derby, a yearly budget) deserve a date in the URL so each year gets its own page, with an evergreen hub above them.
Review within 48 hours. Check Search Console for the URL's Top Stories and Discover impressions and fix the headline, image or hub links for next time.
Watch AI Mode. Adams' observation that AI Mode can replace news links with Wikipedia within days means the archive value of a live blog is lower than it was. A good hub page and a "Highly Cited" original report outlive the live blog.
At DPG Media this is daily practice. Sophie Winkels describes in an interview with Jaap van Zessen (December 2024) how the SEO editors check Google Trends every day, prepare structured plans and dossiers for big news events, and use live blogs for events such as elections.
Should you publish AI-generated content?
Short answer: use AI for support tasks, not for the journalism, and disclose what you do. The evidence, split by type:
Facts: what Google and the rater guidelines say
Google's February 2023 position: "Appropriate use of AI or automation is not against our guidelines" and "Using AI doesn't give content any special gains. It's just content." The same post says "publishers that appear in Google News should use bylines and author information."
The March 2024 scaled content abuse policy applies "no matter whether content is produced through automation, human efforts, or some combination". The spam policies name "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example.
The rater guidelines (section 4.6.6, page 42) require the Lowest rating "if all or almost all of the MC on the page... is copied, paraphrased, embedded, auto or AI generated, or reposted from other sources with little to no effort, little to no originality, and little to no added value". The same section balances it: "the use of Generative AI tools alone does not determine the level of effort or Page Quality rating." Section 4.5.3 flags "AI generated content with made up 'author' profiles" as deceptive.
Since 1 October 2026 Google's gen-AI guide says manual fact-checking is "critical" and references these rater sections directly.
Il Foglio (March 2025): a month-long fully AI-written edition as a deliberate experiment. Editor Claudio Cerasa's own conclusion: articles written by humans "are better, because they always have something more".
Business Insider (2025): 21% of staff cut, reporters allowed to draft with AI, and a "Business Insider AI" author page the union protested. BI's answer to search loss was to raise the share of original reporting from about 40% to over 80% of output, per Nieman Lab.
Facts: readers and rankings
Reuters Institute, Generative AI and News Report 2025: comfort with fully AI-generated news 12%, with a human in the loop 21%, human-led with some AI help 43%, entirely human-made 62%. Only 33% think journalists check AI output before publishing.
Ahrefs (July 2025, 600,000 pages): no correlation between the share of AI text on a page and its ranking. AI as such is not penalised.
Lily Ray (May 2026, 220+ sites that publicly use AI content platforms): 54% lost at least 30% of peak organic traffic, 39% at least 50%. "The trouble starts when the goal becomes volume."
How Dutch newsrooms handle it
De Volkskrant has the strictest rule: no journalism generated by AI, with transcription, translation and research allowed. DPG Media's guideline is "AI versterkt, de mens regisseert" (AI strengthens, people direct), with mandatory disclosure, an internal "koppenmaker" headline generator where editors keep the final say, and some newsrooms that refuse AI-generated text unless AI is the subject, according to Businesswise (July 2026). NRC's opinion desk requires every sentence to be written by the author. NPO, DPG Media, Mediahuis and Talpa agreed joint AI labelling guidelines in September 2026, in line with article 50 of the EU AI Act. The Guardian, AP and the BBC follow the same pattern: AI for assistance, humans for the story, labels for anything significant.
Our verdict: AI-written news articles fail on three fronts at once: rater quality criteria, reader trust, and brand differentiation. Barry Adams' list of good uses is the right scope: alt text, captions, internal link suggestions, transcription, translation, tagging, and research. If your news site "is interchangeable with others in your topical area", as he puts it, AI-written text makes it more interchangeable, not less.
The Search Quality Rater Guidelines, September 2025 edition
The current Search Quality Evaluator Guidelines are dated 11 September 2025 and run 182 pages. As of 7 October 2026 Google still hosts that edition; claims of a 2026 revision are not reflected in the PDF. The change log (page 182) lists for September 2025: updated YMYL definitions, more examples, and minor fixes. The January 2025 edition had added the first definition of generative AI and the spam sections on scaled content abuse and site reputation abuse. John Mueller reminded everyone in May 2026 that the guidelines "are not a guide for search ranking". They do show what Google wants its systems to reward.
Page 21: Quality of the Main Content (section 3.2)
The section opens: "The quality of the Main Content (MC) is one of the most important considerations for PQ rating." The test: "For most pages, the quality of the MC can be determined by the amount of effort, originality, and talent or skill that went into the creation of the content. For informational pages and pages on YMYL topics, accuracy and consistency with well established expert consensus is important." The four criteria, quoted:
Effort: "Consider the extent to which a human being actively worked to create satisfying content." Running "existing freely available content through existing translation software without any oversight, manual curation, etc., would not be considered to have effort."
Originality: "Consider the extent to which the content offers unique, original content that is not available on other websites. If other websites have similar content, consider whether the page is the original source."
Talent or skill: "Consider the extent to which the content is created with enough talent and skill to provide a satisfying experience for people who visit the page."
Accuracy: "For informational pages, consider the extent to which the content is factually accurate. For pages on YMYL topics, consider the extent to which the content is accurate and consistent with well-established expert consensus."
Page 20 adds that "the title of the page is considered part of the MC", and that "the presence or absence of Ads alone is not a consideration for PQ rating."
What the guidelines say about news
Highest quality for news (page 78) is "original reporting that provides information that would not otherwise have been known had the article not revealed it... Very high quality news content will include a description of primary sources and other original reporting referenced during the content creation process." This is the same signal as Google's "Highly Cited" badge and the news topic authority system.
E-E-A-T (pages 26 to 28): "The most important member at the center of the E-E-A-T family is Trust." Trust means "accurate, honest, safe, and reliable". "Untrustworthy pages have low E-E-A-T no matter how Experienced, Expert, or Authoritative they may seem."
YMYL now includes "Government, Civics & Society": election and voting information and "issues of public interest, trust in public institutions". The YMYL table (page 12) uses news examples: news about ongoing violence is clearly YMYL, news about a car accident may be, a high-school basketball game is not.
Reputation (pages 22 to 25): "Prestigious awards or a history of high-quality original reporting are strong evidence of positive reputation for news websites." Raters research the underlying company, not only the domain. For journalists they expect "formal reputation information".
Syndication is not copying (page 42): "Licensed or syndicated content should not be considered as 'copied'... Examples of syndicated content in the U.S. include news articles by AP or Reuters."
Site reputation abuse (page 41): coupons from a white-label service on a news site are abuse; wire content, syndicated news and editorial columns are not. A sponsored "article" that "could be mistaken for an article written by the newspaper" is Lowest.
Freshness (section 18.0, page 158): "Breaking news" queries such as a tornado "Assume users need the information right away"; stale results "should be considered 'stale' and given low Needs Met ratings". Page Quality is different: "some highly reputable news websites maintain 'archival' content" and stale pages can still be High quality.
AI Overviews are not named in the text, but the September 2025 edition added three rating examples whose screenshots show an AI Overview (pages 131, 134 and 137). They are rated Moderately Meets, Slightly Meets and Fails to Meet. None is rated Highly Meets. Marie Haynes' takeaway in "Google quality raters now rate AI Overviews": pages "should offer far more value than a searcher can get simply from the AI answer".
E-E-A-T in the newsroom: Sophie Winkels at brightonSEO
Sophie Winkels, News SEO Specialist at DPG Media, presents "Headlines to bottom lines: turning newsroom E-E-A-T into corporate conversion" at brightonSEO on Thursday 8 October 2026 at 11:20 in Auditorium 2, in the E-E-A-T session alongside Sarah McDowell, Christian Rigg and Katie Mishner. She shared the text of the talk with SearchConsole.nl before the conference. The summary below follows her slides; the quotes are hers.
Why news SEO is different
DPG Media is the largest publisher in the Netherlands and Belgium, with about 80 brands, of which almost 20 are national and regional news sites and magazines, among them Algemeen Dagblad, NU.nl and Het Laatste Nieuws. Sophie came from B2C and e-commerce SEO and "had to unlearn and relearn a lot". Her numbers explain why:
A typical newsroom has about 100 journalists working at the same time.
One news site publishes around 180 unique articles a day. Across DPG's 16 news sites that is more than 2,800 articles every day.
The average lifespan of a news article is 48 hours. Most articles disappear from the search results after that, except category pages and multi-day live blogs.
News is a Your Money or Your Life category, so the stakes are high.
Her conclusion: "You simply can't manually optimize 3,000 pages a day, and not every journalist is an SEO expert." So the goal is not a one-off conversion but readers who return every day, optimisation has to happen before an article is published, and E-E-A-T "has to be baked directly into the site's DNA and template structure". Without that foundation, "you can optimize content all day long, but it simply won't rank" during big breaking news events.
E-E-A-T at DPG Media, with the Eurovision Song Contest as the example
Experience means "boots on the ground" reporting, with visible proof for Google and for readers. For Eurovision the reporter's name and photo sit at the top of the article, the byline is up front, and the text states that the reporter is at the event. The same reporters appear on camera, with the newsroom's logo on the microphone. Original photography beats stock images; where there are no own photos, the desk designs custom visuals in the brand's identity.
Expertise is domain knowledge made visible. A reporter who has covered Eurovision for five years curates the list of performances worth watching and the ones to skip. Internal links connect the latest story to that reporter's earlier coverage of the same topic. When the reporter is not a specialist on a sub-topic, they interview an external expert, credit them, attribute the quotes clearly and link to the expert's official website.
Authority comes from connecting entities to specific authors. A big event is split into breaking news, follow-ups, analysis and evergreens, clustered in a topic hub or "dossier". Instead of "Written by the Entertainment Desk", the reporter on the ground is the named author of every story in the cluster. Every reporter has an author page with a rich bio: speciality, journalistic experience, awards, and an index of everything they published.
Trust is "earned through total editorial transparency": labels when an article is updated, corrected or syndicated, a weekly column in which the editors-in-chief explain editorial choices, and annual newsroom events where readers meet the journalists and ask questions.
What this delivers during a major event, in her words: a bump in traffic compared with competitors who do not prove physical presence, readers who return more often when reporters are put front and centre, and "algorithm immunity", a foundation that core updates do not wipe out. That last claim matches Barry Adams' observation above that sites which lose a core update rarely recover, and Google's February 2026 Discover update language about expertise on a topic-by-topic basis.
The transfer to non-news sites
The second half of the talk translates each principle to e-commerce, affiliate and service sites:
Experience: show proof of the work. Put a real face to the reviewer, add unboxing videos, first impressions and side-by-side comparisons. Add an "Our Pick" label and explain why. For services, show real photos and videos of the team at work with a quote from the colleague doing it.
Expertise: spotlight the team, their client cases and their credentials, invite external experts, and give every expert an author page that links to their articles and profiles.
Authority: timing and topical focus. Publish as a topic starts trending in Google Trends, reuse evergreen content when the topic returns, and attach one experienced colleague as the go-to author on every page about that subject.
Trust: involve customers. Ask for a product photo with every review, run a live Q&A with the CEO, or let leadership explain business decisions in a blog post.
The business case she puts on the final slides: brand awareness, because people see your experts more often; returning visitors, because they know who you are; stronger loyalty; and resilience against future core updates. Her four takeaways: "Show who you are as a brand and as real people. Show, don't just tell. Always show the receipts, bring the proof. Look outside your industry for fresh inspiration."
Her earlier public work shows the foundations. In July 2023 she presented "SEO voor nieuwssites" at the SEO Benelux Summer Meetup, covering Top Stories, the News tab, Discover, E-E-A-T ("news falls under Your Money or Your Life") and NewsArticle, LiveBlogPosting and isAccessibleForFree markup. In 2024 she discussed SEO and AI on the Jerrel Arkes podcast, recorded at brightonSEO, and in November 2024 she and the author of this article explained how DPG Media's SEO editors worked the US elections: daily Google Trends monitoring, dossiers and live blogs, an internal "Chat-DPG" tool, and caution about full AI automation. Her tip for publishers on SEO Brein: start with the who, what and where of the story, "the name of a politician, Valentine's Day, or the village where the news happens".
What newsroom E-E-A-T looks like in practice, combining her material, the rater guidelines and Google's documentation:
Bylines on every article, author pages with Person and ProfilePage markup, credentials, beats and links to the author's other profiles. Barry Adams' question "How trusted are your journalists?" is a Discover question as much as a Search question.
Original reporting with named primary sources, so the piece can earn "Highly Cited" and the Highest rating. Sophie's "show the receipts" is the same rule from the newsroom side.
Corrections policy, contact information, ownership information and clear labels for opinion, updated, corrected and sponsored content, as the Google News and Discover policies require.
Topic-level depth in dossiers with a named reporter. The February 2026 Discover update and the news topic authority system both judge expertise per topic, not per domain.
Brand. Preferred Sources, the Subscribed label in AI Overviews and the Follow feature in Search profiles all reward publishers readers choose by name.
How news works in ChatGPT and Claude
Both assistants answer news questions by searching the live web and citing sources. Neither sends much traffic. The Reuters Institute Digital News Report 2026 puts weekly chatbot use for news at 10% (17% of 18 to 24 year olds), and only 4% of people say they "always or often" click through to the source, against 19% from search engines. Chartbeat measured AI sources at under 1% of pageviews despite 200% growth.
ChatGPT
Search: ChatGPT search launched on 31 October 2024, with inline links and a sources panel, and opened to users without an account in February 2025. OpenAI partnered with news and data providers for weather, sports, stocks and news. A May 2026 change that made brand names clickable raised referrals sharply, but most of those clicks land on homepages, per Similarweb data reported by Similarweb.
Crawlers (OpenAI bots documentation): OAI-SearchBot indexes sites for ChatGPT search; "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers." GPTBot is for training only. ChatGPT-User fetches pages when a user asks, and OpenAI says "robots.txt rules may not apply" because the action is user-initiated. OpenAI revised this documentation in December 2025 to say the search and training bots share crawl results.
Deals: Press Gazette's deals and lawsuits tracker lists OpenAI agreements with AP (July 2023), Axel Springer (December 2023), Le Monde and Prisa, the Financial Times, Dotdash Meredith, News Corp (more than $250 million over five years), The Atlantic, Vox Media, Time, Condé Nast, Hearst, Future, Axios, Schibsted, The Guardian, The Washington Post, Folha and, in September 2026, the Times of India and the Indian Express. Tollbit data shows licensed publishers get a click-through rate almost seven times higher in ChatGPT. Ads arrived in ChatGPT in January 2026 without a revenue share for publishers.
Lawsuits: The New York Times case (filed December 2023) survived a motion to dismiss and is in summary judgment briefing in the Southern District of New York as of September 2026. Ziff Davis, the Intercept, Alden's newspapers, Danish publishers, Britannica and others have sued too.
Search: Claude added web search on 20 March 2025, worldwide for all users by May 2025. Claude decides when to search; recent events and news are explicit triggers. In the API web search tool, "citations are always enabled for web search", each with the URL, title and a quoted passage, and developers "must" show them to end users. Current models have a knowledge cutoff of June 2026, so news answers come from live search, not training data.
Crawlers (Anthropic support article, updated April 2026): ClaudeBot collects content for training, Claude-User fetches pages when a user asks, and Claude-SearchBot indexes pages "to improve search result quality". Anthropic says all three respect robots.txt and Crawl-delay, and that blocking Claude-SearchBot "may reduce your site's visibility and accuracy in user search results". Search Engine Journal contrasted this with OpenAI's and Perplexity's user-triggered fetchers, which may ignore robots.txt.
Deals: none with news publishers in the Press Gazette tracker. Adweek noted the pattern: never signed a publisher licensing deal, never sued by a publisher. Anthropic did settle the authors' class action Bartz v. Anthropic for $1.5 billion (final approval July 2026) over pirated books used in training, and faces Reddit's scraping suit in California state court.
Share: Similarweb put Claude at about 9% of generative AI web visits in May 2026, the fastest growing of the group, behind ChatGPT (53%) and Gemini (about 27%).
What to do
Allow OAI-SearchBot, Claude-SearchBot, Claude-User and ChatGPT-User if you want to be cited. Decide separately on GPTBot, ClaudeBot and Google-Extended, which only affect training. Blocking training bots does not remove you from any search product.
Make sure the article content is in the raw HTML. Arnout Hellemans' advice on the expert page applies: compare rendered and raw HTML, because many AI crawlers do not render JavaScript.
Measure citations, not just clicks. John Shehata's three-layer framework (January 2026) puts crawler access first, brand prompts second, and citations on trending queries third, and argues AI Overviews deserve more attention than standalone chatbots for news.
Brand is the moat. Muck Rack found generative models cite journalism in 49% of recency-related answers, and Reuters and Axios most of all: wire-grade accuracy and fast, structured reporting get cited.
News SEO tools: NewzDash and the alternatives
News SEO needs minute-level data. Generic rank trackers refresh daily, which is useless for Top Stories. Note the spelling: the tool is NewzDash, with a z. "News Dashboard" is the competing Trisolute product.
Founded by John Shehata, formerly Global VP Audience Development at Condé Nast and head of search at Disney ABC. Tracks Top Stories, Google News, organic and video carousels for trending queries every 15 to 60 minutes in 40+ countries; News Trends with estimated hourly volumes; Content Autopilot for publish-to-index times; DiscoverPulse for real-time Discover monitoring from a user panel; and since June 2025 an AI Overview tracker for trending news queries. Three plans (Lite, Niche, Enterprise), prices on request. Clients named on the site include BBC, CNN, Forbes and NBC News. Shehata also runs the News and Editorial SEO Summit (NESS), a 3,000-member News SEO Slack, and publishes studies such as 2026 News SEO Trends & Predictions and the full news SEO tool stack for 2026.
German, since 2012. Rank tracking every 15 minutes for Top Stories, live blogs, publisher and video carousels and Google News in about 15 countries, with Missing Topics and Index Rate reports. Publishes yearly visibility reviews that include AI Overview share on news SERPs.
Real-time audience analytics and headline testing. Estimates Discover as a separate referrer and integrates Search Console. Its network data is the industry benchmark quoted throughout this article.
Publisher analytics with pre-built Discover and SEO dashboards; its community published the analysis of Discover's shift to AI summaries, YouTube and X.
Dutch editorial analytics (Nijmegen): real-time dashboards, notifications, headline A/B tests and the User Needs model. Adjacent to SEO rather than a rank tracker.
Google Search Console
Free and authoritative: the Discover report, the Google News report, the News search type filter, and since 2026 the generative AI reports for Search and Discover. Impressions only for the AI reports; AI Mode has counted in the totals since June 2025.
Trending Now for story selection. Sophie Winkels: "use Google Trends every day."
Semrush, Ahrefs, SISTRIX, Conductor
General suites with news-relevant features: Semrush tracks Top Stories as a SERP feature and sells an AI Visibility Toolkit; Ahrefs' Brand Radar covers AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity and Copilot; SISTRIX's Visibility Index is the common benchmark after core updates; Conductor (which absorbed Searchmetrics in 2023) has an AI search module.
AI visibility tools
Profound (enterprise), Peec AI (Berlin), Otterly.AI and ZipTie track citations and share of voice in ChatGPT, AI Overviews, AI Mode, Perplexity, Gemini and Claude. None is news-specific, and all work from prompt sets you define.
Did Aleyda Solis write about news SEO?
Not as a dedicated topic. Aleyda Solis' AI search work (SEOFOMO, LearningSEO.io, LearningAIsearch.com, her SEO Summit Brasil deck from November 2025) is aimed at brands, ecommerce and SaaS. Her weekly SEOFOMO does track the publisher debate closely, and she occasionally comments. On Google's May 2026 promise of more links in AI answers she wrote that "more AI answers links don't automatically mean more traffic", and on Google's AI optimisation guide that "SEO principles still apply in AI Search". She was also one of the twenty experts in NewzDash's 2026 predictions survey. For news-specific guidance, Barry Adams, John Shehata, Lily Ray and the WTF is SEO? newsletter are the primary sources.
Reading list: Nieman Lab and Press Gazette
Two outlets cover this beat with data. The pieces that shaped this article:
Predictions for Journalism 2026: "Sorry, the explainer is dead" names the losers (explainers, how-tos, evergreens, recipes) and the winners (hyperlocal, breaking news, scoops, first-person, investigations).
Protect the base: fast server responses, article content in raw HTML, semantic HTML, one URL per story, indexable pagination, a clean news sitemap, and lean NewsArticle markup with three image sizes.
Optimise before publishing: headline with entities, a summary intro, a 16:9 image over 1200 pixels, og:title and og:image set deliberately, max-image-preview:large.
Build topic authority on purpose: story hubs, topic pages linked from the first mention, sustained coverage per beat, and authors with real profiles.
Treat Discover as a bonus: report it separately, never let it set the editorial agenda, and expect core updates to swing it.
Win breaking news: the live-blog routine above, plus a wrap-up that can become "Highly Cited".
Keep AI out of the bylines: AI for transcription, translation, tagging, alt text and headline options, with human fact-checking and disclosure. No scaled pages.
Set your crawler policy: allow the search bots of OpenAI and Anthropic if you want citations; decide on training bots separately; use the Search Console generative AI control only after testing per section, because excluded sites also lose Top Stories placements inside AI Overviews.
Measure what matters: Search Console's Discover, News and generative AI reports, a news rank tracker for Top Stories, citation tracking for AI Overviews, and, above all, direct and subscription traffic. Chartbeat's data shows internal and direct traffic are the only growing sources.
Ask for the follow: add the Preferred Sources button, claim your Search profile, and push newsletters and apps. Barry Adams: "The worst thing you can be as a publisher is forgettable."
Frequently asked questions
Do AI Overviews show up for news searches?
Rarely. Google said in May 2024 that it aims not to show AI Overviews for hard news topics where freshness and factuality matter. NewzDash measured AI Overviews on 3.47% of trending news queries in the US in 2025, and Ahrefs found them on 6% of news queries in September 2025. Health, science and technology news trigger them far more often than sports, world or entertainment news. Since mid-2026 Google also embeds the Top Stories carousel inside some AI Overviews, which NewzDash saw on 15.5% of US and 17.46% of UK trending news queries that show Top Stories.
How does Google Discover decide what to show?
Discover is a personalised feed. Google says content is automatically eligible once it is indexed and meets the Discover content policies, and that no special markup is needed. The feed matches entities and topics in your articles against a user's interests and activity. Google rewards large 16:9 images of at least 1200 pixels wide (allowed via max-image-preview:large), clear headlines without clickbait, timely and original content, and topic-level expertise. The February 2026 Discover core update explicitly targets sensational content and rewards sites with expertise in a given area.
Should a news publisher use AI-generated content?
Not for the journalism itself. Google does not penalise AI as a method, but its scaled content abuse policy and the rater guidelines treat pages created with little effort, originality or added value as Lowest quality no matter how they were made. Only 12% of people are comfortable with fully AI-generated news, against 62% for human-made news, according to the Reuters Institute. AI is useful for support tasks: transcription, translation, headline suggestions, alt text, tagging, internal link suggestions, and research. Every published AI-assisted sentence needs human fact-checking, and Google now calls that manual review critical.
How should you handle a live blog or a match report that is only relevant for a few hours?
Publish early on one stable URL with a keyword-focused slug, put live or updates in the headline, and update the headline, the top summary and dateModified as the story develops. Google keeps recognised live coverage in its priority crawl queue for about 24 hours. Link the live page from the homepage and from the story hub. After the event, either rewrite the same URL into a wrap-up, or publish a separate wrap-up and link to it from the live page. Keep the news sitemap to articles from the last two days, and never change dates to make old content look fresh.
How do ChatGPT and Claude find and show news?
Both run a live web search for news questions and show citations. ChatGPT search uses OAI-SearchBot to index sites and ChatGPT-User to fetch pages on request; GPTBot only controls training. OpenAI has licensing deals with more than twenty publishers, including AP, Axel Springer, the Financial Times, News Corp, The Guardian and The Washington Post. Claude added web search in March 2025, always shows citations for search results, and uses Claude-SearchBot and Claude-User; ClaudeBot controls training. Anthropic has no news publisher licensing deals as of September 2026. AI chatbots send less than 1% of publisher pageviews, and only 4% of people say they often click through from a chatbot to the source.
What is NewzDash?
NewzDash is a news SEO platform founded by John Shehata, formerly Global VP Audience Development at Condé Nast. It tracks Top Stories, Google News and organic rankings for trending queries every 15 to 60 minutes, monitors Google Discover through its DiscoverPulse panel in 40 countries, and since June 2025 tracks which trending news queries trigger an AI Overview and whether your domain is cited. The main competitor is the Trisolute News Dashboard. Chartbeat, Marfeel and Google Search Console cover the analytics side.
Sources and credits
Research for this article was carried out on 7 October 2026. Thanks to Barry Adams, John Shehata, Lily Ray, Jessie Willms and Shelby Blackley, Marie Haynes, Barry Schwartz, Charlotte Tobitt and the Press Gazette team, Andrew Deck and the Nieman Lab team, whose public work is quoted and linked above. Primary sources by group:
Julian Aijal. Senior SEO Specialist at DPG Media, Julian Aijal merges his front-end development background with technical SEO, web performance, and publisher SEO. He drove significant organic growth at DPG Media and authored key TNW articles on Schema and Knowledge Graphs, focusing on user-centric, high-performing sites. Since 2026 he is also Website Strategist for TEDxAmsterdam and builds AI tools like PvdAI.