SEO News January 2026: Zero Updates, One Big Concession

SEO news January 2026 at a glance: Google confirmed no ranking update at all, conceded under regulatory pressure that sites may eventually opt out of AI Overviews, and made Gemini 3 the model behind those answers worldwide. The detail is below, or jump to the SEO news January 2026 FAQ.

January 2026 is the rare month with nothing in the ranking-update column. Google's status dashboard runs from the December 2025 core update, finished around 29 December, to the Discover update of 5 February. A clean gap.

The month's story happened elsewhere. In one week the UK's competition regulator opened a consultation on forcing Google to let publishers opt out of AI Overviews, Google said it was exploring exactly that, and Gemini 3 became the default behind those Overviews. Whether your content feeds AI answers became a configuration decision.

No confirmed update, six volatility windows

Google confirmed nothing in January. That is a finding, not a gap in the reporting: the official history jumps from the December 2025 core update, which ran 18 days from 11 December, into February.

The trackers said otherwise. Search Engine Roundtable logged spikes on 6, 12, 15–16, 21, 27 and 29–30 January, with thirteen independent trackers showing elevated readings, and called it an "incredibly volatile month".

Six volatility windows in 31 days with nothing confirmed behind them is a false-positive machine for any regression watchdog keyed to date correlation. Widen the comparison window, and require corroboration from several trackers before you flag a client drop; that discipline is what our Race Control module exists to enforce. Some blogs have published a "January 2026 core update" anyway. There was none. Google did clarify one thing on 15 January: favicon display is unaffected by core updates.

Regulators forced the AI opt-out question

On 21 January the Competition and Markets Authority opened a formal consultation on conduct requirements for Google. The proposed obligations are specific: publisher choice over how content appears in AI features, transparency, attribution, and the ability to opt out of AI Overviews and AI Mode. Fines under the DMCC Act reach 10% of turnover.

Google answered on 28 January: it is "exploring updates to our controls to let sites specifically opt out of Search generative AI features" — no specification, no timeline. Today's controls hit AI features and conventional search visibility together, the problem the post named but left unsolved.

Nobody has a framework for the decision this creates. A poll of 350-plus practitioners on 29 January split 33.2% who would block, 41.9% who would not and 24.9% unsure. Instrument per-cluster citation value against click value now; that is what the AI Search Visibility module measures: when the flag ships there is no A/B and no rollback data. Keep the crawler categories straight meanwhile. Hostinger found GPTBot allowed on 12% of sites, down from 84% — but blocking a training crawler removes you from parametric memory, not from ChatGPT's search results. They need separate crawl controls.

Gemini 3 now writes Google's answers

On 27 January Gemini 3 became the default model powering AI Overviews globally, and follow-up questions can now be asked straight from an Overview, carrying context into an AI Mode conversation — "one fluid experience", in VP Robby Stein's words. Complex queries had routed to Gemini 3 Pro since 19 January. Synthesis changed in every market on one day with no ranking event behind it, and the Overview stopped being terminal: each follow-up is a fan-out query off your click path.

Personal Intelligence launched five days earlier: an opt-in connection between AI Mode and a user's Gmail and Google Photos, personalising answers from booking confirmations, itineraries and purchase history. It is a Labs experiment, English and US only, and no rank tracker can see it. Where order confirmations sit in a customer's inbox, visibility rests on data you cannot see.

Presence is not fixed either. Stein said on 9 January that Google drops Overviews on query types where people do not engage and generalises that to similar queries, so a cluster can lose its Overview with no ranking change. Where one fails to generate, a featured snippet takes the slot. And Yahoo's Scout, launched 27 January on Claude with Bing grounding, makes Bing index health the lever for a third answer engine.

Core ranking work is the AI substrate

The month's most useful official statement came on 12 January: Google has "integrated all of the core search kind of signals into these AI experiences". Precision page deployment is therefore the AI-visibility programme, not a separate generative engine optimisation track.

Google's own research agreed. Its SAGE paper, published 26 January, quantified the shortcuts that let a search agent skip deep research: Information Co-Location at 35%, where several answers sit in one document, then Overly Specific Questions at 31%, Multi-query Collapse at 21% and Superficial Complexity at 13%. Agents settle on the page that also answers the neighbouring questions — support for interlinked cluster hubs, and a match for Google's 9 January warning against bite-sized chunks. Consolidate the topic cluster onto fewer pages.

One smaller item points the same way. llms.txt failed its first controlled test — ten sites, 90 days either side, eight unchanged and one down 19.7%, and the two that gained traced their gains to PR coverage and new templates. Ship it as hygiene, never as an intervention. Microsoft's engine-authored AEO guide on 22 January put the effort elsewhere: product, offer and review markup, and feed-to-page consistency.

Measurement got structurally worse

On 21 January John Mueller confirmed that Google treats an AI Overview as a single position. Every link inside shares it, so a URL in both the Overview and the blue links generates one Search Console impression, not two. Inferring citation rates by differencing impressions is structurally wrong, which makes an external citation feed mandatory inside Opportunity Telemetry.

Third-party data got dearer too. SearchGuard, Google's anti-bot system, and the loss of num=100 put scraping costs up roughly tenfold. Every volatility index you buy is noisier.

Position tracking inside AI answers fared worse still. Six hundred volunteers ran 12 identical prompts about 3,000 times through ChatGPT, Claude and Google's AI. The same list came back twice under 1 time in 100, and in the same order about 1 in 1,000, while brand appearance frequency stayed far more stable. Share of voice across sampled runs is the only defensible AI-visibility KPI, and the right basis for judging LLM visibility tools. It is also the basis our AI visibility audit reports on.

One first-party source arrived: Bing Webmaster Tools opened a closed beta of an AI Performance report on 28 January covering citations by day, cited pages and the grounding queries Microsoft used — the machine's query, not the user's.

Local shrank, recipes came back, checkout moved

On 16 January AI Overview local results began displacing traditional local packs, showing one or two businesses instead of three and stripping call buttons from many listings. Joy Hawkins of Sterling Sky documented "50% declines or more" in local visibility, with AI local packs on about 8% of keywords and rising. For a multi-location client that is the month's largest structural loss, and it lands on Google Business Profile visibility.

AI recommendation does not backfill those slots: SOCi's study of 350,000 locations found rates of 1.2% in ChatGPT, 7.4% in Perplexity and 11% in Gemini against 35.9% in Google's 3-pack. Recommended locations averaged 4.1 to 4.3 stars against 3.9 — a threshold, not a graded score.

Recipes moved the other way on 22 January: carousels returned, Overviews began showing key ingredients rather than the recipe, and on branded searches the publisher's page outranked a demoted Overview. Answer depth is tunable and reversible.

Commerce moved inside the interface. The Universal Commerce Protocol launched on 11 January with Shopify, Etsy, Wayfair and Target, carrying agents from discovery to post-sale support. Three days later Google barred merchants from showing higher prices on Google than on their own site, policed by suspension and by monitoring that adds items to carts. On a Shopify build, feed-to-site price drift is now a suspension risk, and a purchase can complete with no session.

The traffic numbers disagree, and that is the finding

Two flagship studies landed days apart with opposite framings. Graphite, using Similarweb data across 40,000-plus large US websites, found organic search traffic down only 2.5% year on year, the top ten sites growing 1.6%, and the sharpest declines in the top 100 to 10,000 — where most enterprise sites sit. It put AI Overviews on 30% of SERPs, cutting click-through 35%.

Datos and SparkToro, reading clickstream data across tens of millions of US users, found Google searches per US desktop user down nearly 20% year on year against 2–3% in Europe. Both hold at once: query volume per user is falling faster than click volume, and the US is degrading far faster than the EU, so forecast them separately. Two caveats belong in either: AI tools are under 1% of US desktop activity, and zero-click rates have plateaued.

Publishers are gloomier than either dataset: the Reuters Institute's 12 January report found respondents expecting search traffic to fall 43% within three years, and Tailwind CSS reportedly cut 75% of its engineers after a 40% Google traffic decline. One number refuses to reconcile: Reuters puts AI Overviews on about 10% of results, Graphite on 30%. Neither is verified, so measure prevalence on the client's keyword set.

The January 2026 timeline

What to watch next

• The opt-out specification. Google set no timeline and the CMA is the forcing function, so expect the controls UK-first.
• The DMA milestones. Preliminary findings fall about three months from 27 January, conclusions about six. Whether AI chatbot providers qualify for Google's search data is the line to watch.
• Bing's grounding queries. January was closed beta; the report reached public preview on 10 February. It is the first first-party view of the machine's rewritten query.
• ChatGPT ad pricing. Reporting on 22 January put OpenAI's ads at roughly a $60 CPM sold per impression, not per click. Value organic AI citations as impression inventory.
• Feed-to-site price drift. Price parity is enforced by suspension and by a bot that adds items to carts. Audit the feed against live prices.

SEO news January 2026: frequently asked questions

January gave the industry a quiet ranking chart and a loud policy one: no confirmed update, a new model behind every AI answer, and the first hint that you may one day choose whether those answers use your content. Earlier issues are in the SEO news archive.