AI search in late 2026: what Google’s latest changes mean for marketers
More links in AI Mode, a first look at Google paying publishers for AI answers, and a long spam update. What changed and how to adjust content and reporting.
By KINETECH10 min read
On this page
Three things changed in Google’s AI search between late August and late September 2026. Search Engine Roundtable’s monthly recap, published on 1 September, reported that AI Mode gained link carousels for developing stories while AI Overviews started steering searchers into AI Mode. Reporting in mid-September put Google’s AI contribution pilot, which pays invited publishers when their content significantly shapes AI answers, in the spotlight. And on 24 September 2026 Google began its fourth spam update of the year, a global rollout expected to take up to two weeks.
For marketers the message is mixed. Google says there is nothing special to do for generative AI responses, yet more of the journey now happens inside AI surfaces, and Google’s spam policies explicitly cover attempts to manipulate generative AI responses. This piece separates what is confirmed from what is speculation, and ends with a practical GEO and SEO playbook.
Key takeaways
- Search Engine Roundtable’s recap of 1 September 2026 reported link carousels for developing stories in AI Mode and AI Overviews pushing users into AI Mode.
- Google’s AI contribution pilot pays invited publishers only when content contributes significantly at the answer-generation stage; early earnings are described as small.
- The September 2026 spam update began on 24 September, applies globally in all languages and may take up to two weeks.
- Google’s spam policies explicitly name attempts to manipulate generative AI responses in Search as spam.
- Google’s guidance remains “nothing special” for AI responses: original, citable, well-structured content and clean technical SEO still do the work.
What changed in Google’s AI search in August and September 2026?
Late August and September 2026 brought three notable shifts: more visible links and a stronger push into AI Mode, a publicised pilot in which Google pays some publishers for content that shapes AI answers, and a global spam update starting on 24 September. Google also repeated that sites need no special optimisation for generative AI responses.
| Change | Status | Source |
|---|---|---|
| Link carousels for developing stories in AI Mode | Reported live | Search Engine Roundtable recap, 1 Sep 2026 |
| AI Overviews pushing searchers into AI Mode; “Show more” button dropped | Reported live | Search Engine Roundtable recap, 1 Sep 2026 |
| Search Console generative AI performance report expands to more users | Reported rollout | Search Engine Roundtable recap, 1 Sep 2026 |
| “Nothing special to do” for generative AI responses; no special ranking preference for Reddit | Google statements | Search Engine Roundtable recap, 1 Sep 2026 |
| AI contribution pilot pays invited publishers | Invite-only pilot | Search Engine Roundtable, Semrush (via Digiday) |
| September 2026 spam update | Began 24 Sep 2026, up to two weeks | Google Search Status Dashboard |
| Gemini 3.8 Live audio models in Search Live | Rolling out | Google’s September AI roundup |
None of these changes rewrites the fundamentals, but together they show where Google is heading: more questions answered inside AI Mode, more explicit links back to sources within those answers, and tighter enforcement against attempts to game them.
How are AI Mode and AI Overviews changing where clicks go?
AI Overviews increasingly hand users on to AI Mode for follow-up questions, so more of a search session can happen inside Google’s conversational interface. At the same time AI Mode is showing more links, including carousels for developing stories. Being cited inside those answers is becoming as important as ranking in the classic results.
Search Engine Roundtable’s monthly recap, published on 1 September 2026, reported that AI Overviews were pushing searchers into AI Mode and had dropped the “Show more” button, and that AI Mode gained link carousels for developing topics. Google’s September roundup, published on 2 October 2026, also confirmed Gemini 3.8 Live audio models rolling out across Search Live, which extends conversational search to voice.
Our analysis: what it means for traffic
- Fewer, better clicks. When an answer resolves the simple part of a question, the visits that remain tend to come from people who want depth, a comparison or a transaction. Judge AI-search traffic on conversion and engagement, not raw volume.
- Citations become a visibility metric. A page that is cited in an AI Mode answer earns brand exposure even when the user does not click. Track it alongside rankings.
- News and fast-moving topics benefit most from the carousels. If you publish timely analysis, clear dates, bylines and up-to-date facts matter more than ever.
- Informational content without original value is most exposed. Definitions and generic how-tos are exactly what AI answers summarise without needing a click.
Is Google paying publishers for AI answers?
Yes, on a small, invite-only scale. Through a pilot run in Search Console, invited publishers can accept terms and earn when their content contributes significantly to responses in the Gemini app, AI Overviews or AI Mode. Only the answer-generation stage counts, and early earnings have been described as small compared with ad revenue.
The pilot was first spotted in April 2026, and Digiday reported on it on 14 September 2026, as covered by Search Engine Roundtable the same day and by Semrush on 22 September. According to that coverage:
| Question | What has been reported |
|---|---|
| Who can join? | Invited publishers only; Google has approached at least dozens, with particular appeal to small and mid-sized publishers and beyond news sites |
| What earns money? | Content that “contributes significantly” to generating an AI response shown to users |
| What does not count? | Pages only linked after an answer is generated, or used afterwards to confirm a fact |
| Which surfaces? | Gemini app, AI Overviews and AI Mode |
| Reporting | A Search Console report showing monthly earnings and payment history |
| Opting out | Possible at any time from the same settings page |
| Terms and calculation | Terms not published; one executive called the calculation “opaque” |
| Early earnings | Described as small compared with ad revenue |
Our take
Do not build a revenue plan around this pilot. It is invite-only, its terms are unpublished and the payout method is undisclosed. Its real significance is the signal: Google is now measuring which pages shape AI answers, which makes original, source-worthy content more valuable than pages that merely restate it.
What did the September 2026 spam update target?
Google released the September 2026 spam update on 24 September 2026. It applies globally to all languages, may take up to two weeks to roll out and is the fourth confirmed spam update of 2026. Google described it as a normal spam update and has not said it targets AI-generated content or AI answer manipulation specifically.
Two details stand out. Search Engine Roundtable noted that the August 2026 spam update finished in two days, while this one was expected to take far longer; Google’s John Mueller said it was likely to take longer than some previous ones. And when we checked Google’s Search Status Dashboard on 6 October 2026, the rollout was still listed without a completion note, so ranking movement may continue for a few more days.
The policy line worth knowing
Whatever this particular update targeted, Google’s spam policies explicitly describe as spam any attempt “to manipulate generative AI responses in Google Search”, alongside classic ranking manipulation. The same policies define scaled content abuse as many pages generated primarily to manipulate rankings rather than help users, whatever tool produced them.
- Mass-produced AI articles that add nothing original are a scaled-content risk, with or without a spam update.
- In our reading, hidden text or other tricks designed to steer AI answers would be hard to square with the generative-AI manipulation clause.
- If traffic dropped after 24 September, review pages against the spam policies before changing anything else, and wait for the rollout to finish before drawing conclusions.
How should you measure visibility in AI search?
Measure three layers: what Search Console reports for generative AI features, whether your brand and pages are cited in AI answers across Google, ChatGPT and Perplexity, and the quality of the visits AI surfaces send. Clicks alone understate AI visibility and overstate the value of thin traffic.
- Turn on Search Console’s generative AI reporting where your property has access, as availability has been expanding.
- Build a prompt panel. List 30 to 50 questions your buyers actually ask and check monthly which sources AI Overviews, AI Mode, ChatGPT and Perplexity cite. Record whether you appear, and who appears instead.
- Segment AI referrals in analytics. Track conversions, engaged sessions and assisted revenue from AI surfaces separately from classic organic search.
- Track brand demand. Rising branded search and direct traffic are often the downstream effect of being cited without a click.
- Annotate update dates. Mark 24 September 2026 and similar rollouts so you do not misattribute spam-update movement to content changes.
What this means for teams building now: a GEO and SEO playbook
Google’s own advice is that there is nothing special to do for generative AI responses, and the evidence supports a focused version of good SEO: original, citable passages, clear entities and authorship, structured data, fast crawlable pages and honest measurement. Avoid shortcuts aimed at AI answers; they now sit squarely inside the spam policies.
Content
- Answer first. Open each section with a direct 40 to 60 word answer, then add depth. Self-contained passages are what AI answers excerpt.
- Add something only you have: first-party data, worked examples, pricing ranges, screenshots, expert commentary. Restated common knowledge is the easiest content to replace.
- Use tables, steps and definitions for comparisons and processes; they are easy to cite accurately.
- Name authors and dates. Bylines, credentials and absolute dates help both readers and systems judge freshness and trust.
- Prune, do not pad. Merge or remove thin pages rather than publishing more of them into a spam-update cycle.
Technical
- Keep pages fast, server-rendered and crawlable, with canonical URLs and clean internal linking.
- Add accurate structured data (Organization, Article, FAQ where genuinely relevant) that matches the visible page.
- Keep entity details consistent across your site, profiles and directories.
Reporting
- Report citations and AI referral quality next to rankings and clicks.
- Treat the AI contribution pilot as an experiment if you are invited, not as a line in the budget.
For a deeper comparison of the two disciplines, read our guide to GEO vs SEO. If you want help auditing how your site appears in AI answers, our AI-powered SEO service covers content, structured data and measurement.
Frequently asked questions
AI Mode is Google’s conversational search experience, and AI Overviews increasingly lead users into it for follow-up questions. It answers more of a query inside Google but still links to sources, including carousels for developing stories. For SEO, being cited within those answers now matters alongside ranking, and the quality of clicks matters more than their volume.
Google says there is nothing special to do for generative AI responses. In practice, pages that get cited tend to be crawlable, fast and well structured, with direct answers, original data or expertise, clear authorship and accurate structured data. Avoid tactics designed to manipulate AI answers: Google’s spam policies explicitly treat them as spam.
Google called it a normal spam update. It began on 24 September 2026, applies globally in all languages and may take up to two weeks. Google did not say it targets AI content specifically, but its spam policies cover scaled content abuse and attempts to manipulate generative AI responses in Search, so thin mass-produced pages are a risk.
Only through an invite-only pilot. Invited publishers accept terms in Search Console and earn when their content contributes significantly to responses in the Gemini app, AI Overviews or AI Mode at the generation stage. Pages merely linked or used to confirm facts do not count. Early earnings have been described as small compared with ad revenue.
Start with Search Console’s generative AI performance reporting, which has been expanding to more users. Then segment AI referrals in your analytics, run a monthly panel of buyer questions to see which sources AI answers cite, and watch branded search and direct traffic, which often rise when you are cited without a click.
Google’s spam policies target purpose, not tools: scaled content abuse means generating many pages mainly to manipulate rankings rather than help users. AI-assisted content that is original, accurate and useful is not spam by that definition. Mass-produced pages that add nothing new are at risk, especially during spam update rollouts like the one that began on 24 September 2026.
Sources
- September 2026 spam update — Google Search Status Dashboard, 2026-09-24
- Google September 2026 Spam Update Is Rolling Out — Search Engine Roundtable, 2026-09-24
- Spam policies for Google web search — Google Search Central, 2026-08-28
- Google AI Contribution Pilot - Paying Publishers For Use Of Content In AI — Search Engine Roundtable, 2026-09-14
- Google Tests Paying Publishers for AI Answers — Semrush, 2026-09-22
- September 2026 Google Webmaster Report: Spam Update, AI Mode Takeover, GoTo Redirects & More — Search Engine Roundtable, 2026-09-01
- The latest AI news we announced in September 2026 — Google, 2026-10-02
Published by KINETECH, a digital product and AI development company in Pune, India. More guides
Want to know how AI search sees you?
We audit how your pages appear in AI Overviews, AI Mode and other answer engines, and fix the content, structure and measurement gaps.