Your Next Customer Might Ask an AI First — How to Get Found in AI Search

Nearly half of consumers now ask AI tools for local business recommendations. That is up from 6% a year earlier. Those answers get built from ordinary search results, listings and reviews. That's the honest answer to how to get found in AI search. Keep accurate business information. Keep current reviews. Publish pages that answer real questions in plain language. Google published its own guidance in May 2026 saying the same thing.

What is AI search, in plain terms?

Inside Google, AI Overviews and AI Mode write a summary at the top of the results. Outside Google, ChatGPT, Copilot, Gemini and Perplexity answer from a mix of training data and live retrieval from the web. Voice assistants are the same thing with a speaker.

Your customer doesn't experience any of it as "AI search". They ask for a good florist, dentist, gym or repair shop near Barrie. They read two sentences and call one of the names in them. Being in those names is the whole question. The industry calls the work generative engine optimization (GEO). In practice, it's the plain-writing part of SEO.

Do your customers actually use this to pick a business?

Yes, and faster than most owners assume. BrightLocal found in March 2026 that 45% of consumers now use AI tools for local business recommendations, up from 6%. That puts AI third behind Google and Facebook. It also puts AI ahead of Yelp and Tripadvisor (BrightLocal AI trust study). 88% of consumers fact-check what the AI told them. Usually that means reading your reviews.

That's fragmentation, not replacement. In BrightLocal's Local Consumer Review Survey 2026, Google's share as a place people read reviews fell from 83% to 71%. Apple Maps nearly doubled, from 14% to 27% (BrightLocal LCRS 2026).

What Google itself says you should do

Google published official guidance on 15 May 2026. It reads as a rejection of most of what's sold as AI SEO. Its position is simple. The AI features are rooted in core Search ranking and quality systems. They work through retrieval over Google's index plus "query fan-out". The guide still says "the best practices for SEO continue to be relevant". Its strongest line covers unique, valuable, non-commodity content. That work "will likely influence your website's presence in generative AI search in the long run more than any of the other suggestions in this guide" (Google Search Central). Read that before you pay anyone for this.

The tactic Google says you don't need

The `llms.txt` file is supposed to describe your pages to AI crawlers. You are told to drop it at your site root. It's the default "core step" in nearly every AI-search listicle this year. Google's guide says machine-readable files like it aren't needed to appear in its generative AI features.

What the data says about it in practice

Ahrefs analysed 137,000 domains. 28% of them publish an `llms.txt`, and 97% of those files received zero traffic in May 2026. 96% of the requests that did arrive came from bots. No AI bot requested a file that didn't exist (Ahrefs). Spend that hour on the pages it was supposed to help.

Why your clicks dropped even though your rankings didn't

Pew Research measured 900 US adults' browsing in March 2025. When an AI summary appeared, users clicked a traditional result in 8% of searches. That compares with 15% when no summary appeared. 1% clicked a link inside the summary. 26% abandoned browsing after a page with a summary, versus 16% without (Pew Research).

Google disputes the framing. It calls the study's methodology flawed and its queryset skewed. It says it "consistently direct[s] billions of clicks to websites daily". It also says it hasn't seen significant drops in aggregate traffic (Search Engine Land). Both can be true. Aggregate traffic can hold while a single informational search gets fewer clicks. The AI answer is now the top of your funnel. Either you're named in it, or someone else is.

Does a page for every wording of the question help?

No, and it's the fastest way to look like everyone else. Mass-producing pages mainly to manipulate rankings is what Google's spam policies call scaled content abuse (Google spam policies). It also contradicts the guide's line that unique, non-commodity content is what counts. Thirty thin pages for every wording of one question don't get you cited more often. One page that answers the question properly does.

Which searches get an AI answer, and which don't

About 18% of Google searches in Pew's March 2025 sample triggered an AI Overview. The summaries averaged 67 words, and Wikipedia, YouTube and Reddit were the most-cited sources, holding 15% of the links.

Query length is the pattern that matters. Pew found 8% of one- or two-word searches produced a summary. That rose to 53% for searches of ten words or more (Search Engine Land). Semrush's analysis of 10 million keywords put AI Overviews on 13.14% of US desktop searches in March, up from 6.49% in January (Semrush). Long, conversational questions are where the AI answers live. It's how someone asks for a salon open late, a clinic taking new patients or a repair shop near Barrie. Test your own pages against those.

What actually gets content cited

A KDD 2024 paper on generative engine optimization measured visibility gains of up to 40% from optimisation techniques (arXiv). The techniques that worked are dull: be specific, include real numbers, quote sources, and answer the question in the first sentence.

How to get found in AI search: three things to fix this week

One – your business information, complete and identical everywhere

Your legal name, service area, phone, hours and the services you actually offer. Put the same details on your Google Business Profile, your site, your social pages and every directory you're listed in. AI answers get details wrong when your own details disagree. An old phone number on a directory is why an assistant sends someone nowhere. get your business information right everywhere

Two – current, answered reviews

45% of consumers consult AI about local businesses. Reviews are what those answers lean on hardest. A profile whose newest review is three years old gives an AI nothing to say about you today. Ask every customer the same way. Answer every review, including the bad one. reviews are what AI reads about you

Three – a website that answers the questions customers actually ask

Not a brochure with "quality service" on it. Pages should say what you do, what you turn away, what area you cover and how fast you respond. Those answers are the material an AI can quote. a website that answers real questions

And only after those three, structure and schema

Schema is the genuine part of what vendors oversell: it helps machines read your services, hours, service area and location accurately. It's a finishing touch on correct content, not a fix for absent content. If a page doesn't answer a real question, structured data just labels nothing.

What doesn't work, and who's selling it

• `llms.txt` as a magic file, when Google says it isn't needed and 97% of those files were never fetched.
• A page for every query variant, which is scaled content abuse.
• Link packages and "entity" schemes sold as AI visibility, with no way to show you a result.
• Separate "GEO programmes" billed on top of the SEO you already pay for, when the deliverables are the same work.
• Anyone guaranteeing you'll be "the answer" for a given prompt. Nobody controls what a model says.

Nobody has published a defensible measurement of how much of local service search happens inside an LLM. If someone quotes you a percentage, ask for the study.

How to check whether it's working

Open ChatGPT or Perplexity once a month and ask the three questions your customers ask, in their words, with your town in them. Write down which businesses get named and whether the details about yours are right. Then watch your analytics for referral traffic from chatgpt.com, perplexity.ai and copilot.microsoft.com. It's a small number today. Semrush measured outbound referral traffic from ChatGPT up 206% between January 2025 and January 2026, though over 20% of that goes to Google itself (Semrush). Most answers still come from what the model already learned.

What this means for a local small business

The shortlist an AI builds comes from Google's index, Reddit threads, YouTube, review platforms and local news. That is the open web. If you're not in those places, you're not in the shortlist. That's why AI visibility follows reputation and mentions rather than tricks.

Common questions about AI search

Does SEO still matter if AI answers the question?

Yes, and Google says so in writing: its AI features are built on core Search ranking and quality systems and lean on the open web. What changes is the payoff. Being named in the answer matters more than being clicked, and you get named by being specific on your own pages.

Do I need an llms.txt file?

No. Google's May 2026 guide states that machine-readable files like `llms.txt` aren't needed to appear in its AI features. Ahrefs found 97% of the files it studied received zero traffic in a month. Don't pay anyone who calls it the core step.

Will AI replace Google for local searches?

No. BrightLocal's 2026 data shows fragmentation rather than replacement. Google's share as a review destination fell from 83% to 71%. Apple Maps nearly doubled, from 14% to 27%, while AI use rose to 45% from 6%. Consumers now consult around six review sites each.

How long does this take to show up?

Months, and mostly it's ordinary work: reviews, profile accuracy and content that answers real questions. Google's own guide frames the content side as influencing your presence "in the long run". That's a polite way of saying nobody can promise you a date.

Does this cost extra compared with normal SEO?

No. Accurate business information, current reviews and pages that answer questions are the fundamentals of ordinary SEO too. If an agency quotes a separate "GEO package" on top of the retainer for the same deliverables, you're paying twice. Ask what gets built, and in what order.

Do this before you read another AI SEO article

Open an AI assistant and ask it the three questions your customers actually ask. Write down the names it gives you. Then check whether it states your hours and service area correctly. Fix anything wrong at the source. That twenty minutes tells you more than a vendor audit. If you'd rather have the underlying work done, that's our SEO for Google and AI search service. It starts with the boring three: business information, reviews and content. Want to talk it through instead? Book a free consultation.