How to show up when customers ask AI instead of Google

Customers ask assistants who to call now. What decides whether your business gets named, which local SEO work transfers, and what to skip this quarter.

You show up in an AI answer roughly the same way you show up in local search: by being findable, consistent and well reviewed across the sources those systems read. There is no separate AI ranking system to buy your way into. An assistant asked “who should I call for a leaking water heater in South Austin” builds a short list out of web content, maps and business profile data, reviews, and mentions of your business elsewhere. If a business is invisible in local search today, it is usually invisible in the assistant’s answer too, and for the same reasons.

So the useful version of the question is not “what do I do differently for AI.” It is “which parts of the existing work now matter more, and what is being sold to me that does nothing.”

The change is the shape of the answer, not the evidence behind it

Search used to hand back a page of options and let the customer sort them out. An assistant hands back two or three names and a sentence about why. The question arrives in full sentences with conditions attached: open Saturday, works out in Pflugerville, does natural stone rather than pavers. The answer arrives as a recommendation instead of a list.

That compression cuts both ways, hard. Being named in a three-name answer is worth far more per appearance than sitting tenth on a results page. Being left out is worse than being tenth, because nobody scrolls a chat reply looking for alternatives. Fewer people reach your site before deciding, which means more of the decision happens on material you published months ago rather than on the page you would have shown them.

What did not change is where the machine got its information. It came from the same places local search has always drawn on.

Where an assistant gets a local answer

Some assistants search the live web while answering and cite the pages they used. Others answer from what they absorbed in training, which is older and blurrier. Most now do a mix, and the mix shifts with every release. Underneath all of it sits the same material: search indexes, mapping and business profile data, review platforms and the text inside the reviews, directory and trade association listings, and the pages businesses publish about themselves.

The specifics of which system leans on which source, and how heavily, change every few months. That instability is exactly the argument for building on the parts that do not move. A business with accurate profile data, steady reviews and pages that plainly state what it does and where it works has given every one of those systems something correct to repeat. A business without them has given them nothing to work with, and a system filling a gap will fill it with whoever did do the work.

Most of what works here is work you already know

A business that ranks locally, has a complete Google Business Profile, collects real reviews and states its services and service area clearly is already most of the way there. Those are the same four things that decide whether you appear in the map results, and they are the four things a language model can find and repeat about you. There is no second checklist.

The one piece that gets less attention than it deserves is consistency.

Entity clarity is the part most businesses get wrong

Machines assemble a picture of your business from fragments scattered across dozens of sources, and they resolve conflicts by picking whichever version looks strongest. Your job is to stop giving them a conflict to resolve.

If your profile category says drain cleaning, your website navigation says plumbing solutions, and a directory from 2021 lists a disconnected number and a suite you moved out of, you have published three partial businesses. One legal name spelled one way. One phone number. One address format. One set of service names used on the site, on the profile, and in every listing you can still log into. It is unglamorous work and it is the highest-value hour on this list.

What deserves more emphasis than it used to

Most of it is work you would recognize from ordinary SEO, done more literally. AI assistants quote pages that answer a question in plain sentences, so the practices below have moved from nice-to-have to the whole game.

Write the answer directly under the question

Extraction systems lift the sentence that answers the thing. A page that spends three paragraphs warming up before it gets to the number gives them nothing liftable, so they take the competitor page that opened with it. Put the direct answer in the first two sentences under the heading, then add the nuance, the exceptions and the reasoning for the person who wants them. Readers prefer this too, which is why it is worth doing regardless of what any model does next year.

Be the source for your trade’s local questions

The questions customers ask assistants are the ones they used to ask you on the phone. What does this cost around here. Do I need a permit in this county. How long does it take. Which material holds up in this climate. How do I tell if it needs replacing or repairing.

Generic national answers to those already exist in abundance. What is scarce is a specific, local, experienced answer with a named business behind it. Publishing those is the practical version of being citable, and it is the clearest current reason to own pages rather than run on a profile alone, a question we took apart in do you need a website if you have a Google Business Profile.

Reviews are the raw material for “best in town” answers

When somebody asks for the best roofer in a suburb, the system is working from text that describes local businesses, and reviews are by far the largest body of customer-written description that exists about companies like yours. Volume, recency and what the text actually says all feed that picture, which is a second reason to run a steady collection habit instead of an annual burst. The ranking side of it is covered separately in what reviews actually do to your local ranking.

Structured data helps the parsing, not the persuasion

Marking up your pages with LocalBusiness schema, hours, services and areas served removes guesswork about who published a page and which business it belongs to. It promises nothing and it wins no rankings. It is cheap, you do it once, and it stops a machine having to infer facts you could simply have stated.

What is not worth paying for

Every shift in search produces a product category selling access to it, and this one is no exception. Before you sign anything sold as AI-search optimization, hold it against this list:

• AI SEO packages promising placement in answers. No slot exists to sell. Nobody can buy your name into an assistant’s reply, and anyone promising a ranking in AI answers is promising something they do not control.
• Copy written for machines. Keyword stuffing fails the same way it failed in 2011, except now the reading system works from meaning, so it fails harder while also reading badly to humans.
• Submissions to new AI directories. Most appeared in the last two years, carry no traffic and are read by nothing.
• Anyone quoting you a percentage of your customers now coming from AI. Nobody has that figure for your business. We do not either, and the honest answer is below.

Check what the assistants actually say about you

This takes fifteen minutes and it is the only diagnostic that matters. Open two or three assistants and ask, in the phrasing a customer would use, three things: who are the best plumbers in your town, tell me about your business by name, and does your business do the specific service you want to be known for.

Read the answers for what is wrong rather than for whether you appear. Wrong hours, an old phone number, a service you dropped, an area you no longer cover, a competitor described as doing your specialty. Then fix each one at the source the machine read: your site, your profile, the directory that still holds the old number. There is no correction line to call, and changes land on the systems’ own refresh schedules, which vary and are not published. Fix it and check again next month.

The parts nobody can measure yet

Attribution here is genuinely murky, and it would be dishonest to say otherwise. A visit that started with an assistant recommendation often arrives with no useful referrer, or lands as a direct visit, or shows up days later as somebody searching your business name. Referral reporting from these platforms is early and inconsistent between them. So the effect of doing this well tends to appear as more people contacting you already knowing who you are, which is real and hard to put in a report.

The practical response is old fashioned. Ask on the phone how they heard about you, record the answer on every lead, and read the pattern quarterly. That log is worth more right now than any dashboard claiming to measure AI traffic.

What to actually do this quarter

A quarter is enough to work this list top to bottom without hiring anyone, and the early items feed the later ones.

1. Fix entity consistency first. One name, one number, one address, one set of service names, everywhere you appear.
2. Complete the Business Profile properly, including services and service area, since it feeds both the map results and the assistants.
3. Make review collection part of finishing a job rather than a campaign.
4. Publish four to six pages that answer real customer questions for your trade and your town, each stating the answer in its opening sentences.
5. Add structured data to your site once.
6. Ask the assistants about your business monthly and log what they get wrong.

None of that is speculative and all of it pays off if AI answers turn out to matter less than the loudest predictions. That is the test for anything in this category: if the tactic only makes sense in a world where the predictions are right, skip it.

If you would rather this ran as a system, with the checking and the correcting on a schedule instead of a good intention, that is what AI implementation covers.