Ask how AI is changing SEO and the honest answer has three parts. Google now answers many searches with an AI summary above the links, and fewer of those searches end in a click.
People also search inside chatbots, which read the web and cite a few sources. And visibility is now counted inside those answers, not only in rankings.
What has not changed is the raw material: Google builds its AI answers from the same index and ranking systems as its ordinary results, so AI search visibility still starts with pages that can be crawled, indexed and quoted.
What changed in the last year
Most of what a site owner can now see and control in AI search arrived in the last twelve months. The dates below are Google’s own.
| When | What Google launched | What it changed |
|---|---|---|
| January 2026 | Work on a control that lets sites opt out of AI features in Search, as the UK’s competition regulator consults | Site owners get a say in AI answers |
| May 2026 | A guide to optimizing for generative AI features; spam policies extended to AI responses; FAQ rich results gone from 7 May | Google’s own list of AEO and GEO tactics to ignore |
| May 2026 | Preferred sources in AI Overviews and AI Mode | Searchers can ask to see the sites they choose more prominently |
| June 2026 | Generative AI reports and an AI opt-out control in Search Console, tested first in the UK; more links inside AI answers | AI impressions get a report of their own |
| July 2026 | AI Overviews and AI Mode brought into one Search experience | One AI search instead of two |
| August 2026 | The reports and the opt-out control reach every site worldwide on 31 August | Every site can measure its AI visibility, and choose it |
AI answers sit above the results, and publishers see fewer clicks
Google said at its developer conference in May 2026 that AI Overviews has over 2.5 billion monthly active users, and that AI Mode passed 1 billion “in just a year”.
Publishers measured what followed. For the Reuters Institute’s January 2026 report on journalism trends, Chartbeat counted Google organic search traffic to over 2,500 news sites: down by a third globally between November 2024 and November 2025, and by 38% in the US.
The report is careful to say it is not clear how much of that fall is down to AI Overviews. The news leaders it surveyed still expect traffic from search engines to almost halve over the next three years.
Google’s answer is more links inside the answers. In June 2026 it said it had increased the number of inline links in AI responses and added website previews “to encourage people to click through”.
And searchers read the summaries. In Pew’s February 2026 survey of 5,119 US adults, 60% said they do.
US adults who read the AI summaries at the top of search results
Search itself has not shrunk. In July 2026 Alphabet said AI Mode, like AI Overviews, “is driving an incremental increase in Search queries overall”.
Both can be true: more questions get asked on Google, and more of them are answered on Google’s page. That is why ranking and being cited are now separate jobs.
Search now happens outside Google too
Chatbots are a second front door, but not yet the main one.
In the same Pew survey, about half of US adults said they do not use chatbots at all, so for most people the AI they meet is still the one inside Google. B2B buyers are further along.
Forrester’s survey of nearly 18,000 business buyers, published in January 2026, found 94% use AI during their buying process, and that because AI search tools can deliver “incomplete or unreliable information”, buyers check what they read with peers, product experts and industry analysts.
That changes what search has to do for a B2B company. A buyer may never reach your page; they read an answer that names a few vendors, then check those names with people they trust.
Being named depends on what the web says about you as well as on your own site. Google’s guide says its AI features “can show what’s being said about products and services across the web, including in blogs, videos, and forum discussions.”
What Google says has not changed
Asked whether SEO still matters for AI search, Google’s guide to optimizing for generative AI features, updated in July 2026, answers “In short, yes!” Its AI features are “rooted in our core Search ranking and quality systems”, and optimizing for them “is optimizing for the search experience, and thus still SEO.”
Still required
- Indexed, with a snippet. A page “must be indexed and eligible to be shown in Google Search with a snippet” to appear in AI Overviews or AI Mode.
- Crawlable, with nothing in robots.txt, the CDN or the hosting blocking Googlebot.
- Content a summary cannot replace. Google calls it “non-commodity content”: a first-hand review offers a view of its own, while a summary of existing content “simply restates information already available elsewhere.”
- Structured data that matches the page, and important content in text rather than only in images.
What you can ignore
Google’s guide to generative AI features also lists the AEO and GEO tactics Google Search does not reward:
- llms.txt and other AI text files, which Google Search ignores, though an llms.txt file can still guide AI agents outside Google.
- Chunking. “There’s no requirement to break your content into tiny pieces for AI to better understand it.”
- Writing for AI. “You don’t need to write in a specific way just for generative AI search”, or chase every long-tail variation.
- Special schema. Structured data “isn’t required for generative AI search, and there’s no special schema.org markup you need to add”, though it still earns rich results.
- Manufactured mentions. Google says “seeking inauthentic ‘mentions’ across the web isn’t as helpful as it might seem.”
- A page for every variation. Pages built for each fan-out query “primarily to manipulate rankings or generative AI responses” break the scaled content abuse policy.
AI made average content cheap
Google has never banned AI writing. Its guidance on generative AI content, updated on 1 October 2026, says AI “can be particularly useful when researching a topic, and to add structure to original content”. What Google penalizes is volume without value.
Its spam policies call that scaled content abuse: many pages generated “for the primary purpose of manipulating search rankings and not helping users”, “no matter how it’s created”.
The policy now gives as an example “Using generative AI tools or other similar tools to generate many pages without adding value for users”, and in May 2026 Google made clear that it covers generative AI responses in Search too.
The same guidance explains why a person still has to check the work: generative models “don’t retrieve facts, but predict a likely sequence of words”, so every AI-written fact, title, meta description and piece of structured data needs a manual check before it goes live.
The bigger change is competitive. Now that AI-written marketing copy is everywhere, a tidy summary of what other pages say is exactly the commodity content Google’s guide warns about. What stands out is what a model cannot produce: first-hand experience, original data and a clear point of view.
More crawlers read your site, and you choose which
Each AI company runs its own crawlers, and most separate the one that feeds search answers from the one that gathers training data. That split lets a site stay visible in AI search while opting out of training.
| Crawler | Owner | What it does | If you block it |
|---|---|---|---|
| Googlebot | Search, including AI Overviews and AI Mode | Pages stop being crawled for Search | |
| Google-Extended | A robots.txt token for Gemini training and grounding outside Search | Google Search is unaffected | |
| OAI-SearchBot | OpenAI | ChatGPT search | Pages are left out of ChatGPT search answers |
| GPTBot | OpenAI | Collects training data | Content is marked as not for training |
| Claude-SearchBot | Anthropic | Claude’s search results | Anthropic says visibility in its search may fall |
| ClaudeBot | Anthropic | Collects training data | Future content is excluded from training |
| PerplexityBot | Perplexity | Perplexity’s search results, not training | Perplexity advises allowing it to appear in results |
ChatGPT-User, Claude-User and Perplexity-User are different again: they fetch a page because a person asked about it, and OpenAI notes that for its own, robots.txt rules “may not apply”.
Two things shut AI crawlers out without anyone deciding to.
Scripts are the first. Google renders JavaScript, and its own changelog notes it has done so “for multiple years now”. The other AI crawlers’ documentation makes no such promise, so put the text that matters in the HTML the server sends.
Network defaults are the second. Since 15 September 2026, Cloudflare’s default settings block “mixed-use” crawlers, those that blend search with training or agents, on pages that carry ads, for new customers, new sites and every free account, TechCrunch reported in July.
This site lets them all in, on purpose. Its robots.txt names nineteen search and AI user agents in one group with the wildcard, because a crawler follows only the most specific group that names it: give one a group of its own and it stops seeing the rules written for everyone else.
AI visibility now has its own reports
- Google Search Console has generative AI performance reports. They show impressions in AI features such as AI Overviews and AI Mode, by page, country, device and date, and the same impressions still count in the main Performance report.
- Bing Webmaster Tools has had an AI Performance report in public preview since February 2026. It counts citations across Microsoft Copilot and AI-generated summaries in Bing, with the pages cited and the grounding queries behind them.
- Google Analytics can give visits from ChatGPT, Gemini, Copilot, Claude and Perplexity a channel of their own. Google’s help page supplies the pattern for an “AI assistants” channel and says to place it above Referrals.
None of these shows what an assistant actually said about you, and Google adds that “No third-party tool has access to our internal ranking or AI systems.” Asking the assistants the same set of buyer questions every month, and recording who gets named, fills that gap.
The SEO work changed too
The other half of the question is what AI does to the job. Research, briefs, audits and first drafts now run through AI tools, and on this site through a coding agent that edits the site’s files under a written rulebook.
This article was researched that way: the agent pulled search suggestions, read the pages that rank, and opened every source quoted here in a browser to match the wording.
What AI cannot do is be right on its own.
MIT researchers ran 24,000 searches in 243 countries in 2024 and 2025, comparing AI search results with traditional ones, and published in February 2026. The AI results linked to the long tail of the web significantly less, and drew on more low credibility sources.
In two experiments in the same paper, AI search tools halved the time people spent on a task, but people over-relied on them when the model was wrong.
That is why Google insists on a manual check, and it is where the human part of SEO now sits: deciding what is true, what is worth saying, and what only your company knows.
Where to start
Measure first, because the Search Console and Bing reports cost nothing and show which pages already appear. Fix access next, since a blocked crawler cancels everything after it.
Then rework the pages behind the questions your best-fit buyers actually ask, with facts and experience a summary cannot copy.
Mentions come last only because they take longest: analysts, industry experts and peers are who B2B buyers trust to confirm an AI answer, so expert commentary placed where buyers look carries further than another post on your own blog.
● About the author
Mark Zides is the founder and CEO of Katama. He founded it to do the work rather than sell it. Meet the team.