SEO for publishers in AI search
When ChatGPT or Google AI Overviews answer a question with your reporting, your publication should be the source they name. SEO for publishers requires articles to rank in Google News and Top Stories and AI assistants to cite and link your pages when they summarise the stories you broke. Start by measuring where you stand today with a free AI visibility report. It shows which prompts on your beats cite your pages and which cite a competitor.
AI visibility for publishers
AI visibility for a publisher is how often AI assistants name its articles and cite them with links when they answer questions on its beats. It's a separate result from ranking. An article can sit in position 1 on Google while the AI Overview above it quotes a rival. To get your publication cited in AI search answers, start with access.
Your robots.txt has to let OpenAI's search crawler (OAI-SearchBot) fetch your pages, or ChatGPT search can't use them. Publish facts nobody else has, such as original numbers and on-the-ground detail, with named sources. An answer engine can rebuild a rewrite of agency copy from 40 other sites, but it has to cite the outlet that holds the only copy of a fact.
Check your crawler rules first, then list the stories from the last quarter that no other outlet could have written. Those stories are the ones an AI assistant has a reason to credit to you.
AI visibility for a publisher is how often AI assistants name its articles and cite them with links when they answer questions on its beats.
How AI summaries cut traffic
AI summaries cut traffic because they answer the query on the results page, so fewer readers click through. Pew Research Center tracked 68,879 searches by 900 US adults and found users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when none did. They clicked a link inside the summary in 1% of visits. For news queries, Similarweb data showed searches with no click to a news site rose from 56% to nearly 69% between May 2024 and May 2025.
The losses aren't spread evenly. In the Reuters Institute's 2026 report, Google search traffic to publishers had already fallen 33% globally, and publishers expect it to drop 43% over three years. According to an AP summary of the report, lifestyle and utility content such as weather and horoscopes saw the steepest declines, while breaking news held up better. Carly Steven, who set up the Daily Mail's editorial SEO team, put it plainly: "AI Overviews has exacerbated an existing problem and challenge that we have, but it isn't the root cause of all of our search challenges."
So the strategy that protects traffic while ChatGPT answers directly is to invest in reporting a summary can't replace. Publishers in the Reuters Institute survey plan more original investigations, with a +91 point net balance, and less service journalism. Before you cut anything, check which of your sections depend on queries that now trigger an AI answer.
AI summaries cut traffic because they answer the query on the results page, so fewer readers click through.
Citation tracking for media
A publisher tracks when AI summarises its articles by running a fixed set of prompts through each AI engine on a schedule and logging which URLs each answer cites. Server logs show that a bot fetched your page. They don't show whether the answer credited you.
Pull the top 50 to 100 queries for each beat from Google Search Console and rewrite them as the questions readers type into ChatGPT.
Run each prompt in ChatGPT and in Google with AI Overviews, weekly for news beats and monthly for evergreen ones.
Log every cited URL, and note whether the answer names your publication or credits the story to someone else.
Compare your share of citations per prompt with your three closest competitors.
Attribution is the step teams skip, and it goes wrong. Columbia Journalism Review's Tow Center tested eight AI search tools on excerpts from real articles and found they gave wrong answers about the source in more than 60% of queries. DeepSeek credited the wrong source 115 times out of 200. If you syndicate your stories, check whether the citation goes to your URL or to the partner's copy.
Content structure AI prefers
AI engines quote self-contained passages that state a fact plainly under a heading that matches the question. Answer engines pull out chunks of a page. A paragraph that only makes sense after the three before it rarely gets quoted. Put the news and its source in the first two sentences. Give numbers with their attribution in the same sentence. Write subheads as the question a reader would ask.
Dates and authorship matter more for news than for any other content. Mark up articles with NewsArticle schema that specifies datePublished and dateModified. Add the author to that markup. Update dateModified when a story changes. Paywalled stories need the isAccessibleForFree property so Google can tell gated text from cloaking. Live blogs are the hardest format to get quoted. Add a short summary at the top that you rewrite as the story develops.
AI engines quote self-contained passages that state a fact plainly under a heading that matches the question.
Brand entity work
Entity work makes your publication one recognisable thing to machines, so AI engines can identify it and attribute claims to it. Language models link facts to entities. If your masthead reads "The Courier" in one place and "Courier Media Group" in another, a model can treat them as two sources or fold you into a namesake. That's one route to the misattribution the Tow Center measured.
Snoika aligns brand facts across your site as one of three layers of entity work, so the About page and footer give the same name and founding details. It creates or fixes the publication's Wikidata entity. It adds schema markup to every page, so each article points back to the same Organization. Check whether your Wikidata item exists and whether your sameAs links point to your official social profiles.
Entity work makes your publication one recognisable thing to machines, so AI engines can identify it and attribute claims to it.
Google News fundamentals
Google News still rewards fast indexing and clear dates. Visible bylines also matter. Google considers sites for Google News automatically, so eligibility depends on those signals. A news sitemap speeds up discovery. Google's documentation says it should list only articles created in the last two days, with up to 1,000 news tags per sitemap.
Use one date per article page. Pages that show several dates, such as related-story timestamps, can make Google pick the wrong one. Author pages with real biographies help Google's systems and AI engines alike. Check Search Console each week for sitemap errors and for articles indexed hours after they went live.
Google News still rewards fast indexing and clear dates.
Measuring referral loss
You measure referral loss by comparing click-through rates on the same queries before and after AI summaries appeared. Total traffic is the wrong measure, because it combines AI effects with changes driven by news cycles and seasonality in one line.
Which baseline to compare against
Export Search Console data for the same months a year apart. Split queries into those that now trigger an AI Overview and those that don't, then compare CTR at the same average position. Ahrefs found that an AI Overview lowered the position 1 click-through rate by 34.5% on average across 300,000 keywords. Use that as a benchmark. A larger gap on your own queries points to content that a summary fully replaces.
Export Search Console data for the same months a year apart.
Where AI referrals show up in analytics
ChatGPT adds utm_source=chatgpt.com to its outbound links, so in Google Analytics 4 (GA4) you can filter sessions by that source. Perplexity and Gemini add these parameters less often, so some of their visits land in Direct. Before you call a dip lost demand, check whether Direct rose by the same amount.
ChatGPT adds utm_source=chatgpt.com to its outbound links, so in Google Analytics 4 (GA4) you can filter sessions by that source.
Snoika for publisher teams
Snoika runs audits and AI visibility tracking in one platform. The same platform handles content planning and publishing. For publishers, its tracking records which of your pages ChatGPT and Google AI Overviews cite for each prompt. That's the per-prompt log described above, kept up without staff hours. The same team runs the platform and the services, with human review on top. The table maps each problem on this page to the step that handles it.
| Publisher problem | Snoika step |
|---|---|
| Clicks falling on AI Overview queries | Audit of affected sections and pages |
| No view of AI citations | Per-prompt tracking in ChatGPT and AI Overviews |
| Coverage gaps a rival owns | Researched content map |
| Misattributed stories | Off-site strategies and entity fixes |
| Board wants proof | Reporting |
Publishers weigh Snoika against tracking-only tools or a manual spreadsheet:
If you only need numbers, a tracking tool is the cheaper choice. If your newsroom has no one free to act on those numbers, a platform that also plans and publishes the fixes is the better fit.
How to get started in publishing
Establish a baseline by recording the prompts on your three biggest beats and who gets cited for each today. Without it, you can't tell whether a new structure or a Wikidata fix changed anything. Run it before you reorganise sections or cut service journalism, because the citation data shows that a section you planned to drop still earns AI mentions. Snoika audits publisher SEO and reports on results, with tracking based on that baseline. Get the free report to see your citations per prompt in ChatGPT and Google AI Overviews.
Establish a baseline by recording the prompts on your three biggest beats and who gets cited for each today.
Trusted by
FAQ
Allow AI search crawlers to access your articles and publish original reporting with named sources. Put the main fact and its attribution together near the top. These steps support citation eligibility, but they don't guarantee that an answer will credit your publication.
Run a fixed set of reader questions through AI search tools and record the URLs each answer cites. Check weekly for news beats or monthly for evergreen coverage. Server logs show crawler visits, but they don't establish whether an answer used or credited your reporting.
Compare click-through rates for the same queries at similar ranking positions before and after AI summaries appeared. Use matching months a year apart, separating queries with summaries from those without. Total traffic alone can't distinguish AI effects from seasonal changes or news cycles.
Prioritise original reporting that gives readers a reason to visit beyond the summary. A practical approach to seo for publishers starts by identifying sections exposed to AI answers. Use query-level click data alongside citation records before deciding which coverage to reduce.
Manual tracking costs staff time rather than a dedicated tracking subscription. The workload depends on how many prompts you check and how often you repeat them. Include time to verify attribution, especially when answers link to syndicated copies instead of your original articles.