AI SEO for Media and Publishers
Publishers are hit twice by AI search: fewer clicks and a new fight to be cited. Chartbeat data reported by the Reuters Institute shows Google organic traffic to news sites fell 33% globally in the year to November 2025. 2 The response is to build direct audiences, publish original reporting AI must cite, set clear crawler rules and measure by brand and subscribers, not clicks alone.
- Google search traffic to 2,576 sites fell 33% globally and 38% in the US between November 2024 and 2025. 2
- Clicks drop when an AI summary appears: 8% of visits against 15% in a US study. 1
- Google-Extended does not affect Search inclusion, so it is not a way to leave AI Overviews. 6
- Retrieval is not citation: in one ChatGPT study, news URLs were cited 12.01% of the times they were retrieved. 10
- We have no published media results. What transfers is the method, proven on local commerce and a personal brand. 9
How are publishers affected by Google AI Overviews?
They lose clicks, and the size of the loss depends on the study and the site. In Pew's analysis of 900 US adults in March 2025 (a general US sample, not news readers), people clicked a traditional result in 8% of visits to pages with an AI summary against 15% without one. 1 Only 1% of visits clicked a link inside the summary. 1
Ahrefs estimates, by comparing December 2025 with a forecast from December 2023, that an AI Overview correlates with a 58% lower average click-through rate for the top result. 3 It is a modelled estimate and does not separate AI Overviews from other SERP changes. Treat it as a direction, not a forecast for your site.
Google's own documentation says it has seen clicks from pages with AI Overviews tend to be higher quality. 5 It gives no figures, so treat it as a claim, not a measurement.
- Search is a major traffic source, and Chartbeat data shows it fell 33% globally in a year. 2
- AI summaries reduce clicks: 8% against 15% in a US study. 1
- Being cited is a second contest that most publishers lose often: news was cited 12.01% of the time it was retrieved. 10
- Licensing income is uncertain, and most leaders expect only a minor contribution. 2
What does the evidence say about traffic, and what does it not say?
The Reuters Institute reports that the Chartbeat decline covers 2,576 sites, including 797 US and 1,031 European ones. 2 In its survey of 280 digital leaders in 51 countries, publishers expect search traffic to fall by more than 40% over three years. 2 Around a fifth expect a loss of more than 75%. 2
AI assistants send little traffic so far. The report calls ChatGPT referrals little more than a rounding error and says Google sends about 500 times as many referrals from search. 2 We found no evidence that AI referrals replace lost search traffic at scale.
What no study shows is your own loss. Search Console reports AI feature clicks and impressions inside the Web search type, with no separate AI report. 5 So you must compare query groups, such as question queries against navigational ones, to see the effect yourself.
| Figure | What it measures | Limit |
|---|---|---|
| 33% global fall 2 | Google organic traffic to 2,576 sites, Nov 2024 to Nov 2025 | Cause not split out |
| 58% lower CTR 3 | Top result CTR with an AI Overview, 300,000 keywords | Modelled; correlation |
| 8% vs 15% 1 | Clicks on results with and without a summary, 900 US adults | One month, US only |
| More than 40% fall expected 2 | Publisher leaders' view over three years | A survey of opinion |
Why is being retrieved not the same as being cited?
AI tools fetch many pages and cite only some. In an Ahrefs study of 1.4 million ChatGPT prompts, news URLs were cited 12.01% of the times they were retrieved, while general search results were cited 88.46% of the time. 10 The authors warn that comparing cited and uncited pages in aggregate can mislead, so read it as a pattern.
The same study found that cited pages had titles closer to the prompt than uncited ones, 0.602 against 0.484 on its scale. 10 A headline written as a clear statement of the answer may therefore serve both readers and retrieval.
For publishers, the lesson is to be the primary source. Original data, documents, interviews and explainers that name their evidence are harder for a summary to replace than a rewrite of a wire story.
Which controls can a publisher use over AI crawling and snippets?
Use the right tool for each goal. Google-Extended is a robots.txt token that controls whether crawled content is used for Gemini training and grounding in certain products. 6 Google states it does not affect inclusion in Google Search and is not a ranking signal. 6
To limit what appears in AI features inside Search, Google lists nosnippet, data-nosnippet and max-snippet, which limit what is shown. 5 Noindex removes a page from the index. 5 Each trade-off is real: a stricter snippet limit can also cut your visibility.
Cloudflare's 2025 data shows the imbalance at scale. It reports crawl-to-refer ratios as high as 500,000 to 1 for Anthropic early in the year, settling at 25,000 to 100,000 to 1 later, a peak of 3,700 to 1 for OpenAI in March, and about 3 to 1 for Google in mid-July after roughly 30 to 1 around April. 4 Ratios swing a lot, so read them as snapshots. Our guide on blocking AI training but still appearing in AI search shows the rules.
Should a publisher expect licensing income from AI companies?
Most publishers expect little. In the Reuters Institute survey, 20% expected significant revenue from AI deals, 49% expected a minor contribution and 20% expected no income. 2 Also, 37% saw platforms paying for content as a key growth opportunity. 2
We found no public price list, and deals are private. Do not build a business plan on licensing until you have a signed term sheet.
A practical step is to make your rights position clear. Publish your terms of use, your crawler rules and a contact for licensing. That does not create a deal, but it makes you easy to reach.
What should a publisher publish and change first?
Put weight on the content AI cannot easily replace. Original reporting, datasets, expert interviews, explainers with named authors and video all fit. In the Reuters survey, publishers' net scores (share planning more effort minus share planning less) for 2026 favoured video (+79), audio (+71) and YouTube (+74). 2 Traditional Google SEO scored -25 in the same survey. 2
Show clear authorship and dates. Google's Article guidance asks you to list each author separately with a link to a page that identifies them. 7 Visible dates and markup dates should match and reflect real publication or update. 8
Update explainers when facts change, and say what changed. A date that moves without real edits misleads readers, and Google asks that dates reflect real updates. 8
- A byline and author page for every article. 7
- Evergreen explainers on your core topics, with a visible 'last updated' date. 8
- Original data or documents in a form that can be quoted and linked.
- Newsletter and app sign-up on every page, to own the audience.
- A rights and licensing page that states your crawler rules.
How should a publisher measure success when clicks fall?
Move from visits to outcomes. Track newsletter sign-ups, subscribers, return visits and direct traffic, and treat Search Console clicks as one input. Google itself advises measuring the value of visits, such as sales or sign-ups. 5
Keep a prompt log of 20 topic questions in your beat. Run each three times on two AI engines every month and record whether your brand is cited, the position and the exact claim attributed to you.
Report in context. We do not guarantee any citation or traffic outcome, and the effect varies by topic. Compare topic groups, not the whole site.
What have Chalk Labs and Rahil Jain done in media?
We have not published results for publishers. Do not read this guide as proof of recovered traffic.
What transfers is the method: query research, answer pages, technical readiness and measurement. It was proven on a local trophy shop in Bengaluru, which reached 27,300 Google clicks in 16 months, and on a new personal site that ranked first for one narrow phrase within a month. 9 Both are Chalk Labs' own work. A publisher has a very different revenue model, so expect to adapt it.
Your first 90 days
Days 1-30
- Build a 20-question prompt log for your core beat and run it on two AI engines, three runs each.
- Split Search Console queries into question and navigational groups to see where clicks fell.
- Review robots.txt and snippet controls against your goals.
Days 31-60
- Add author pages and consistent Article markup with visible dates.
- Rewrite five evergreen explainers to lead with a clear answer and cite their evidence.
- Publish a rights and licensing page.
Days 61-90
- Publish one original data piece with downloadable tables.
- Add newsletter prompts to top-traffic pages and measure sign-ups.
- Rerun the prompt log, compare with day one and report by topic.
The first pages to publish
- Author pages for every writer. Clear authorship, as Google's Article guidance asks.
- Editorial standards and corrections. Shows how you check facts and fix errors.
- Explainers on your core topics. Evergreen answers that cite their evidence.
- Original data pieces. Primary sources are harder for a summary to replace.
- Rights, AI and licensing policy. States crawler rules and gives a licensing contact.
- Newsletter landing page. Builds a direct audience that does not depend on search.
- Topic hubs with dated updates. Group coverage so readers and AI find the core piece.
- Video explainers on watch pages. Google wants a dedicated watch page for each video.
Trust signals and risks
What AI looks for
Why Rahil Jain for Media and publishers?
Method applies
We have not published results of our own in Media and publishers yet. What transfers is the method, shown below.
We have not published results for publishers. What transfers is the method (query research, answer pages, technical readiness, measurement), proven on local commerce (Meru Trophy) and a personal brand (shilikajain.com).
Rahil Jain
Rahil Jain is an AI SEO (answer engine and generative engine optimisation) consultant and co-founder of Chalk Labs.
Results we can show
These are Chalk Labs' own search results. They are not a forecast for your site: results depend on your market, competition and starting point.
How to read these numbers
Chalk Labs reports them from Google Search Console data. We cannot publish the raw console. The shilikajain.com site belongs to Shilika Jain, a Chalk Labs co-founder, so treat it as a demonstration on our own network and not as an independent client result. Meru Trophy is a client of Chalk Labs.
Frequently asked questions
Can we opt out of Google AI Overviews without leaving search?
Do AI Overviews really cut publisher clicks?
Is Search Console enough to track AI Overview traffic?
Only partly. Google includes AI feature clicks and impressions in the Web search type and does not report them separately. 5 You can compare question-style queries against others, and add a monthly prompt log to see whether you are cited.
Will licensing deals replace lost search traffic?
We found no evidence they will for most publishers. In the Reuters Institute survey, 20% expected significant revenue from AI deals and 20% expected none. 2 Deals are private, so treat income as unproven until you hold a signed term sheet.
How long before changes show results?
Nobody can say for your site, and we do not guarantee any citation or traffic result. Our own cases are search-led and measured in months. Compare topic groups quarterly, and read changes over several months before you draw conclusions.
Sources
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research CenterOpened October 20, 2026. Supports: 900 US adults, March 2025: 8% vs 15% clicked a traditional result; 1% clicked a link in the summary; 26% vs 16% ended the session.
- Journalism, media, and technology trends and predictions 2026 · Reuters Institute for the Study of Journalism, University of OxfordOpened October 20, 2026. Supports: Chartbeat decline of 33% global and 38% US; 280-leader survey; AI deal expectations; 2026 priorities; ChatGPT referral scale.
- AI Overviews Reduce Clicks by 58% (re-run of the 34.5% study) · AhrefsOpened October 20, 2026. Supports: 300,000 keywords; 58% lower average CTR for the top result, December 2023 vs December 2025; correlation and caveats.
- Cloudflare Radar 2025 Year in Review · CloudflareOpened October 20, 2026. Supports: Crawl-to-refer ratios for Anthropic, OpenAI and Google, 1 January to 2 December 2025.
- AI features and your website · Google Search CentralOpened October 20, 2026. Supports: Snippet controls, reporting under Web search type, click-quality statement, indexing requirement.
- Google's common crawlers · Google Search CentralOpened October 20, 2026. Supports: Google-Extended controls Gemini training and grounding and does not affect Search inclusion or ranking.
- Article structured data · Google Search CentralOpened October 20, 2026. Supports: List each author separately with a url or sameAs link; datePublished and dateModified in ISO 8601.
- Publication dates in Google Search · Google Search CentralOpened October 20, 2026. Supports: Visible dates, consistent markup, dates must reflect real publication or update.
- Chalk Labs case studies: Meru Trophy and shilikajain.com · Chalk LabsOpened October 20, 2026. Supports: Meru Trophy: 27,300 clicks in 16 months; shilikajain.com ranked first for one narrow phrase within a month. Chalk Labs' own work.
- Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts) · AhrefsOpened October 20, 2026. Supports: Citation rates by retrieved type (news 12.01%, search 88.46%); title similarity 0.602 vs 0.484; authors' caveat about aggregate comparisons.