A customer in Grand Baie asks ChatGPT which hotel suits a family of 4 in December. A finance director in Ebene asks Perplexity for 3 firms that handle payroll across borders. Neither of them scrolls a page of links. They read one answer, and the businesses named in it get the call. If your business is not named, you were not in the running.
That is the change in one paragraph. AI search visibility means being one of the businesses named in that answer. The acronyms around it mostly get in the way, keywords still matter, and AI agents will take all of this further over the next 2 years.
How search changed
For 20 years search worked one way. You earned a place on page one, someone clicked, and you had a visit. Search engine optimisation was the craft of earning that place, and rankings measured success.
That model has changed. Google now puts a summary written by AI at the top of many results, and offers AI Mode for longer, conversational questions. ChatGPT, Perplexity, Microsoft Copilot and Gemini answer questions directly and show a handful of sources alongside. The person gets an answer first and a link second, if they want one.
The effect on clicks is measurable. In a study of the browsing of US adults in March 2025, the Pew Research Center found that users who saw an AI summary clicked a traditional search result in 8% of visits, against 15% for users who did not see one. Links inside the summary itself were clicked in just 1% of visits. We looked at what a B2B site can do about that in when AI Overviews shrink organic traffic.
| Google search visit | Clicked a traditional result |
|---|---|
| An AI summary appeared | 8% of visits |
| No AI summary appeared | 15% of visits |
| Clicked a link inside the AI summary | 1% of visits |
The study does not say search is dead or that websites no longer matter. It says the answer now does much of the work the click used to do. So the question a marketing head should ask has changed too. It used to be “where do we rank?” It is now “when someone asks, are we named, are we described correctly, and can they act on it?”
This is a bigger shift for some sectors than others. A hotel, a clinic or a law firm lives on questions that start with “which” and “who”. Those are the questions an AI answer can settle in one paragraph. These posts go deeper: when AI Overviews shrink organic traffic, branded search and AI Overviews and content systems that work as AI reshapes discovery.
AEO, GEO and LLMO, and why we call it AI search visibility
The industry has produced several acronyms for this work. Answer engine optimisation (AEO). Generative engine optimisation (GEO). Large language model optimisation (LLMO). AI SEO. Each one has its champions and its slide decks.
We think the labels obscure more than they explain. They describe the same goal from slightly different angles, and they invite a board to buy a new discipline when what it needs is a clear view of one outcome. That outcome is visibility, meaning whether your business shows up, accurately, in the answers people now rely on.
So we use the plain name. AI search visibility is how present, accurate and actionable your business is in answers generated by AI. AI search visibility optimisation is the work of improving it. It covers Google’s AI features, the chat assistants and the agents that act on their answers, and it sits on top of good SEO rather than replacing it.
The name matters less than the measure. Whatever you call it, 3 questions decide the result.
- Can the AI systems reach your content? If their crawlers are blocked, you are not in the pool of sources.
- Can they understand it? Clear pages, consistent facts and structured data help AI systems describe your business correctly.
- Do they trust it enough to cite it? Independent mentions, reviews and a consistent presence across the web make you a safer source to name.
Access comes first, and it is often the quiet failure
A common gap is also the least glamorous. A site tells AI crawlers to stay out, often without anyone in the business knowing.
The detail matters, because the crawlers do different jobs. OpenAI’s documentation says GPTBot “is used to crawl content that may be used in training our generative AI foundation models”, while OAI-SearchBot “is used to surface websites in search results in ChatGPT’s search features”. It adds that sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers”. Blocking one is a decision about training. Blocking the other is a decision about being found.
Perplexity draws a similar line. It says PerplexityBot “is designed to surface and link websites in search results on Perplexity” and “is not used to crawl content for AI foundation models”. Anthropic says disabling Claude-SearchBot “prevents our system from indexing” a site for search.
Google is different again. Its crawler documentation states that Google-Extended “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal in Google Search”. Opting out of Gemini training does not remove you from Google.
| Crawler | Run by | What its owner says it does |
|---|---|---|
| GPTBot | OpenAI | Crawls content that may be used to train models |
| OAI-SearchBot | OpenAI | Surfaces websites in ChatGPT search |
| PerplexityBot | Perplexity | Surfaces and links websites in Perplexity results, not used for model training |
| Claude-SearchBot | Anthropic | Indexes content to improve Claude’s search results |
| Google-Extended | Controls use for Gemini, with no effect on inclusion in Google Search |
Blocks also arrive through infrastructure. Cloudflare offers a managed robots.txt that adds directives for AI crawlers, generated from the policies a site owner sets, and only when at least one of those policies is set to block. A setting switched on for a good reason, such as protecting content from training, can also shut the door on search assistants if nobody checks the detail. Treat it as a business decision, made deliberately by someone who understands both sides.
Does keyword optimisation still matter?
Yes, but for different reasons, and no longer as the main measure.
Keywords still matter for 2 reasons. The first is demand. Keywords are the best evidence you have of what people want. The words people type show how they describe a need and how many of them want it. That has not changed, and the questions people ask AI can be longer versions of the same intent. The second is that classic results have not disappeared. Plenty of searches still end on a page of links (we covered the marketing tactics that still work separately), and Google’s own SEO starter guide still advises you to think about the words a user might search for to find your content.
What has faded is the idea that ranking for an exact string is the goal. The same guide says that Google does not use the keywords meta tag and that keyword stuffing is against its spam policies. In our view, AI answers work from the meaning of a question rather than its exact words. They draw on pages that state facts clearly, answer the actual question and agree with what the rest of the web says about you.
Our verdict is to use keyword research to understand demand and intent, keep writing naturally for the words your customers use, and stop treating rankings for exact strings as the main measure. What moves AI search visibility is being a clear, consistent, reachable and citable source about your own business. That means who you are, what you offer, where, for whom and on what terms.
What Google says about its own AI features
There is a lot of advice for sale about special markup for AI. Google’s guidance on AI features and your website is more measured. It says that “specific optimization isn’t required for AI Overviews and AI Mode” and that “all existing SEO fundamentals continue to be worthwhile”, starting with allowing crawling in robots.txt. It also says that meeting its requirements does not mean Google will crawl, index or serve a page.
That is useful in 2 ways. It means the foundations you may already have paid for still count. And it means anyone who promises a certain place in AI Overviews is promising something Google itself does not.
Structured data still earns its place, for a simpler reason. It states your facts, such as opening hours, prices, locations and services, in a form machines read without guessing. It does not buy a mention, but it reduces the chance of being described wrongly, and wrong descriptions are a real cost when an AI answer is the first thing a customer reads.
The direction of travel, from answers to agents
The next step is already visible. AI systems are moving from answering questions to acting on them.
In August 2025 Google announced agentic capabilities in AI Mode, starting with finding restaurant reservations and, in its words, “expanding soon to local service appointments and event tickets”. The same announcement extended AI Mode in English to more than 180 new countries and territories. On the commerce side, the Agentic Commerce Protocol describes itself as “an open standard for programmatic commerce flows between buyers, AI agents, and businesses”, so that a compatible agent can complete a purchase while the business stays the merchant of record.
In practice that means a buyer can hand a task to an agent and review the result. For the difference between an agent and a chat, see when to use an AI agent. 4 use cases make it concrete.
B2B vendor shortlists
A procurement lead asks an agent for 5 logistics providers in Mauritius with cold chain capacity and a quote process. The agent reads service pages, compares what it finds and returns a shortlist. Firms whose services are described only in general terms, or whose quote route is hard to find, drop out before a human ever looks.
Hotel and travel booking
A family asks for a beach hotel suitable for small children, with a dated price for 3 nights. An agent needs to reach that price on the hotel’s own site. If rates only load inside a booking engine it cannot read, the agent may settle for an intermediary listing instead, along with the commission that comes with it.
Retail product comparison
A shopper asks which store nearby has a particular appliance in stock this week. Agents compare price, stock and delivery. Product pages that hide price until a store is chosen, or carry no product data, are likely to be skipped.
Local services
Someone needs a clinic that can see a child today, or a plumber on Saturday. The agent looks for hours, services and a way to book. A phone number with no hours, or a booking form that fails on mobile, ends the conversation.
How a business gets prepared
Preparation is mostly about removing friction for a reader that is not human. The things that matter are the ones above. Allow crawlers where you want to be found, state the facts an agent needs clearly on your own pages, keep those facts consistent everywhere else they appear, and give people a path from question to enquiry or booking that works without a person guiding it. None of that is exotic. Most of it is the work of being clear, done properly and checked from the outside.
Outlook for the next 1 to 2 years
Discovery is moving into the AI answer, and the visit is becoming the step that confirms a decision rather than the step that starts it. Agents began with reservations and purchases, and we expect appointments, quotes and repeat orders to follow. That rewards businesses whose facts are consistent everywhere, whose fundamentals are clean enough to adopt whichever agent standards win, and who are visible in both English and French. This is our reading of the direction, not a forecast, and any of it could arrive sooner, later or differently.
Key takeaways
- AI answers now decide which businesses get named, and people click far less when an AI summary appears.
- AEO, GEO and LLMO describe one outcome. Call it AI search visibility and measure whether you are present, accurate and actionable.
- Check crawler access first. Training crawlers and search crawlers are different, and blocking the wrong one costs visibility.
- Keywords still show demand and intent. Rankings for exact strings are no longer the main measure.
- Google says AI Overviews and AI Mode need no special optimisation. SEO fundamentals still apply, and nobody can guarantee a place.
- Agents are starting to book and buy. Sites that let a machine finish a simple task will be chosen more often.
FAQ
Is SEO dead now that AI answers questions directly?
No. Google says SEO fundamentals still apply to AI Overviews and AI Mode, and many searches still end on a page of links. What has changed is the measure. Being named and described correctly in answers now matters alongside rankings.
What is the difference between AEO, GEO and AI search visibility?
They describe the same goal from different angles. We use AI search visibility for the outcome, how present, accurate and actionable a business is in AI answers, and AI search visibility optimisation for the work of improving it.
Should we block AI crawlers to protect our content?
It depends on which crawler and why. Training crawlers and search crawlers are separate. OpenAI, for example, says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. Decide each one deliberately.
Do we need special schema or markup for AI Overviews?
Google says no special optimisation or markup specific to AI is required. Structured data is still worth having because it states your facts clearly, which lowers the risk of being described wrongly.
Can anyone guarantee that our business appears in AI answers?
No. Google states that meeting its requirements does not mean a page will be indexed or served, and the assistants choose their sources themselves. A credible adviser improves your odds and measures the result, but does not promise a placement.
How do we know whether AI search mentions our business today?
Ask the questions your customers ask, in the languages they use, across several assistants, and record who gets named. Then check whether those assistants can reach and read your site. An audit does this systematically and repeats it.
Where to start
The first step is to see your business the way an AI answer sees it. Current Ventures runs an AI search visibility audit that shows what the main assistants say about you, in English and French, and what to fix first. Or start with a free AI search readiness report.
Sources
- Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results (22 July 2025)
- Google Search Central, AI features and your website
- Google Search Central, Google’s common crawlers
- Google Search Central, SEO starter guide
- Google, AI Mode in Search gets new agentic features and expands globally (21 August 2025)
- OpenAI, Overview of OpenAI crawlers
- Perplexity, Perplexity crawlers
- Anthropic, Does Anthropic crawl data from the web
- Cloudflare, Managed robots.txt
- Agentic Commerce Protocol