How to Get Your Business Recommended by AI Search in 2026

Something changed in how customers find local businesses this year, and most small business owners have not been told about it. People are no longer typing “plumber near me” into a search box and scrolling ten blue links. They are asking an assistant, in a full sentence, and taking the answer it gives them.

“Find me a reliable emergency plumber in my area who does same-day callouts and has good reviews.” “Which dentist near me takes new patients and does Invisalign?” “Who should I use for a new roof, and roughly what will it cost?”

ChatGPT, Google’s AI Overviews, Gemini, Perplexity and Claude answer those questions with a short list, usually three to five businesses. If your business is on that list you get the enquiry. If it is not, you are invisible, and you will never see the search that skipped you. There is no impression, no click, no trace in your analytics.

This is the single biggest shift in small business marketing since Google Maps, and the practical fixes are almost entirely on your own website. Here is what is actually happening and what to do about it.

What Actually Changed in 2026

Three things converged this year, and together they moved the point of decision away from search results and onto AI answers.

AI referred traffic exploded. Adobe reported roughly 393 percent growth in AI-referred retail traffic in the first quarter of 2026 alone. Surveys through the year have put the share of consumers using AI somewhere in their buying journey at around 73 percent. That is not early adopters any more, that is the mainstream.

Agents started browsing the web themselves. Google’s Project Mariner and OpenAI’s Operator are AI agents that open real web pages, read them, fill in forms and complete tasks. Operator has been scoring in the high eighties on complex browser task benchmarks. Google used its 2026 Cloud Next event to push an entire agent stack, including the Agent2Agent protocol for agents that talk to each other.

A common plumbing standard arrived. The Model Context Protocol, originally from Anthropic, became the industry standard for connecting AI assistants to real tools and data, passing 400 million monthly SDK downloads and going fully stateless in its 28 July 2026 revision. Google now offers managed MCP servers across its cloud. In plain English: assistants can now reliably reach out and do things, not just talk.

Put those together and you get a customer journey where an assistant researches, compares and shortlists on the customer’s behalf, and sometimes books the appointment too. Your website stopped being just a brochure that humans read. It became a data source that machines evaluate.

How AI Decides Which Business to Recommend

AI assistants are not running a secret ranking algorithm invented from nothing. They lean on signals that are, mercifully, mostly things you can control:

Structured data. Schema markup tells a machine that this is a business, this is its address, these are its services, this is the price range, these are its opening hours. Without it, an assistant has to guess from your page design, and design means nothing to a machine.
Clear, answer-shaped content. Pages that directly answer a question in the first sentence get quoted. Pages that warm up for four paragraphs get skipped.
Consistency across the web. Your name, address and phone number matching everywhere is what builds what the industry calls entity trust. Mismatched details make an assistant less confident about recommending you.
Reviews and their sentiment. Gemini and Perplexity lean heavily on Google Business Profile, Yelp, Trustpilot and similar sources for local recommendations. Not just star count, but what the reviews actually say.
Whether a machine can read the page at all. If your content only appears after heavy scripting, or your site is slow, or you block AI crawlers, you are not in the running.

Notice what is not on that list: how beautiful your website looks. An AI agent never sees your hero image or your animation. It reads your markup. A gorgeous site with no structure loses to a plain one with clean data every single time.

The 9-Point AI Visibility Checklist

1. Put the answer first on every page

Open each service page with one or two sentences that plainly state what you do, where you do it, and roughly what it costs. “We are an emergency plumbing service covering the Dallas metro and the surrounding area, available 24 hours, with callouts from a fixed price.” That sentence is quotable. A vague brand statement is not.

2. Add proper schema markup

At minimum you want LocalBusiness or the more specific type for your trade, Service markup on each service page, FAQPage on your question sections, and Review or AggregateRating if you have genuine reviews. This is the single highest-return technical change most small business sites can make in 2026, and it is invisible to human visitors.

3. Fix your name, address and phone everywhere

Your website, Google Business Profile, directories and social profiles should carry byte-identical details. “Street” on one and “St” on another is enough to reduce confidence. Pick one format and enforce it.

4. Build reviews deliberately, and reply to them

Ask every satisfied customer, on the day, with a direct link. Reply to every review including the bad ones, because assistants read the replies as evidence of how you handle problems. Ten detailed recent reviews beat fifty ancient one-word ones.

5. Give every service and every area its own real page

One page listing eight services and twelve towns cannot be cited for any of them. A page per service, and a genuine page per area you actually serve, with real local detail, gives an assistant something specific to match a specific question against. Real pages, not thin doorway pages with the town name swapped.

6. Answer the questions people actually ask

Cost, timescale, guarantees, call-out charges, whether you are accredited, whether you cover a given zip code. Write them as questions with direct answers. These are exactly the phrasings people put into assistants, and pages that answer them are the pages that get quoted.

7. Make the site fast and readable by machines

Clean semantic HTML, one H1, sensible heading order, real text rather than text baked into images, and fast server responses. Agents work under time limits. A slow page is often simply skipped rather than waited for.

8. Let the right crawlers in, and add an llms.txt

Check that your robots rules are not blocking AI crawlers you actually want visiting. A growing number of sites now publish an llms.txt file, a plain summary of what the business does and which pages matter, written for language models. It costs nothing and removes ambiguity.

9. Make the next step machine-completable

A phone number alone is a dead end for an agent. A clear contact form, an online booking option, a visible email address and structured opening hours mean an assistant can carry a customer all the way to a booking rather than stopping at “you could try calling them”.

What This Looks Like Trade by Trade

Plumbers and heating engineers. Emergency intent dominates. Make availability, response time and call-out pricing explicit and structured. “Same day”, “24 hour”, “no call-out fee” are the phrases assistants match against panic-driven questions.

Dentists and clinics. Whether you are accepting new patients is the single most asked filter, and it is missing from most clinic websites. Add it, keep it current, mark up your treatments individually, and publish honest price ranges.

Roofers, builders and electricians. Accreditations and guarantees are trust signals assistants look for by name. Put them in text and in markup, not only in a badge image, because an image of a logo tells a machine nothing.

Salons, spas and fitness. Bookability wins. Structured opening hours, a service menu with prices, and an online booking link put you ahead of competitors whose only option is a phone call during business hours.

Shops and e-commerce. This is where agentic shopping is furthest along. Google’s Shopping Graph now carries more than 50 billion listings including small independents. Detailed, structured product data with real stock and price information is the entry ticket.

What Not to Do

The advice going around is not all good, so a few honest warnings.

Do not stuff keywords. Repeating “best plumber London cheap plumber London” does not fool a language model, it just makes your page read like spam to the systems deciding whether to quote you. Write for a person, structure for a machine.

Do not mass-publish AI-written filler. Forty generic articles will not make you visible. They dilute your site and give assistants nothing specific to cite. Fewer, genuinely specific pages beat volume every time now.

Do not block every bot in a panic. Some site owners have blanket-blocked AI crawlers and then wondered why they vanished from AI answers. Decide deliberately which crawlers you want.

Do not buy “AI visibility” packages that promise guaranteed placement. Nobody can guarantee an assistant will name you. Anyone selling that is selling something else.

Do not abandon normal SEO. AI answers are largely built on the same underlying index and the same trust signals. Good traditional SEO is the foundation, not a competing strategy.

Where to Start if You Only Have an Afternoon

In order of return for time spent: add or fix your schema markup, rewrite the first two sentences of every service page to answer the question directly, make sure your business details match everywhere, and ask your last ten happy customers for a review. That is a single afternoon and it covers most of the gap for a typical small business site.

If you want to see where you stand today, ask ChatGPT or Gemini the question a customer would ask, with your town in it, and see who gets named. It is an uncomfortable test and an extremely useful one.

How This Works With a Pay Monthly Website

Everything above is standard on the websites we build. Schema markup, clean semantic HTML, fast hosting, real service and location pages, structured contact and enquiry handling and an llms.txt file are all part of the build rather than an upsell, from $39 a month with hosting, SSL, updates and support included.

The practical advantage of a managed pay monthly site is that this stuff keeps moving. The MCP specification changed in July. Schema recommendations shift. Assistants change what they weight. On a one-off build from three years ago, none of that gets updated unless you pay someone each time. On a managed plan it simply gets done.

If you would like to know where your current site stands, get in touch and we will tell you honestly, including if the answer is that your existing site is fine. You can also compare build costs with our free website cost calculator, or see what is included on our pay monthly plans.

Frequently Asked Questions

Is AI search actually replacing Google?
No, and that framing causes bad decisions. Google itself is the biggest AI answer provider through AI Overviews. What is changing is that fewer searches end in a click to a website, so being the business quoted inside the answer matters more than being tenth on a results page.

How do I know if AI assistants are recommending my business?
Ask them. Use the exact phrasing a customer would, including your town, across ChatGPT, Gemini, Perplexity and Google AI Overviews. Also watch for referral traffic from chatgpt.com and perplexity.ai in your analytics, which is now a visible channel in Google Analytics.

Does schema markup really make a difference?
Yes, and it is the change most small sites are missing. Structured data is how a machine knows what your page means rather than guessing from layout. It also feeds rich results in normal search, so the work pays twice.

How long does it take to show up in AI answers?
Faster than traditional rankings, but not instant. Once pages are crawled and indexed, changes can be reflected within weeks. Reputation signals such as reviews and consistent citations take longer, typically a few months.

Do I need a brand new website for this?
Not always. If your current site is fast, crawlable and editable, most of this can be retrofitted. If it is slow, built on a page builder that outputs unreadable markup, or you cannot edit it without paying a developer each time, rebuilding is usually cheaper than fighting it.

What is llms.txt?
A plain text file at the root of your site that summarises what the business does and lists the pages that matter, written for language models rather than humans. It is an emerging convention, not an official standard, but it is free and it removes ambiguity about what your site is for.

Will AI agents book appointments directly?
Increasingly yes. That is exactly what agentic browsing and MCP-connected tools are for. Businesses with a clear online booking or enquiry route will capture those customers. Businesses that only accept phone calls will lose them silently.

Is any of this worth it for a very small local business?
It is arguably worth more for a small business than a large one. AI answers name three to five businesses rather than a page of ten, and they weight specificity and genuine local relevance, which is exactly where a small local operator can beat a national chain.

Key Takeaways

Customers are increasingly getting a shortlist from an AI assistant rather than a page of search results, and businesses that are not on it never see the enquiry they lost.
AI assistants judge structure, clarity, consistency and reviews. They cannot see your design at all.
Schema markup, answer-first copy, matching business details everywhere and real service and location pages are the highest-return fixes.
Make the next step something a machine can complete, not just a phone number.
Traditional SEO still underpins all of it. This is an addition to good practice, not a replacement for it.

The businesses winning this right now are not the ones with the biggest marketing budgets. They are the ones whose websites are clear, structured, fast and honest about what they do and what it costs. That has always been good practice. In 2026 it finally became measurable.

Real sites, real businesses.

From $39 a month. You own it.