Getting your firm named by ChatGPT, Perplexity and Google AI Overviews

Why a local firm suddenly has to care about this

Something changed in how people look for a dentist, a tax advisor or a plumber. They stopped opening ten tabs and started asking one question and taking the answer. That shift happened faster than almost anyone planned for.

The awkward half of that number is what the assistant says back. It answers with a handful of names, and for most local trades it does not have many to choose from.

Read those two together and the position is unusual. Demand for the answer is rising quickly and the supply of firms the engine is willing to name is thin. That is a narrow window in which a small firm can get named for the cost of doing some housekeeping properly, and it will not stay narrow.

What the engine is actually doing when somebody asks for a dentist near them

It is not reading your website and deciding you are good. It is assembling an answer from sources it already holds and already trusts, then checking whether the same business shows up in several of them saying the same thing. Your website is one of those sources and usually not the decisive one.

That single fact reorders everything the industry sells you. If corroboration across trusted third parties is the mechanism, then work done only on your own pages is work done in the least influential place. It is not useless. It is just last.

Which is why the fix list below starts where it does. Everything in the first three positions is about making one description true in more places. Nothing in it requires a developer, and none of it is fast.

The six things to fix, in the order they pay

1. Your Google Business Profile, completely filled in. Correct primary category, real opening hours, service list, service area, photographs that are not stock, and the same phone number you actually answer. This is the single most cited description of a local firm in existence and most of them are half empty.
2. One name, one address, one phone number, everywhere. Write the canonical version down in a document, then work through every place your firm appears and make them match to the character. Old office addresses on a directory from four years ago are the usual culprit, and they are the cheapest thing here to fix.
3. The registers your profession already has. Chamber listings, professional associations, the register that governs your trade. These carry unusual weight because they are hard to fake, and in Germany most firms are already in several of them without ever having checked what they say.
4. The two or three directories your trade actually uses. Not forty. Pick the ones a customer in your sector would genuinely open, fill them in properly, and stop. Volume of listings is not the signal; agreement between listings is.
5. Reviews, in volume and recently. Reviews are third-party text about you, which is exactly the material an engine is assembling from. Recency matters as much as count: twenty reviews from this year read very differently from forty from 2021.
6. Your own website, last. Pages that answer the questions buyers actually ask, in the words they ask them, with your service area and your credentials stated plainly. This is worth doing. It is the sixth thing to do, not the first.

That figure is about humans, and it is quoted here for a reason that is not the obvious one. The number of reviews you need before a person will book you is roughly the number you need before your firm reads as established to anything else looking at it. The two thresholds sit close together, which is convenient: doing this for customers does the other job at the same time.

How to test whether any of it worked, in twenty minutes a month

This is the part nobody sells you, because it is free and it takes twenty minutes. You cannot manage this without a record, and the record is easy to keep. Do the same thing on the same day each month and write down what happened.

1. Write down five questions a real customer would ask. Not your marketing language. The actual sentence: best dentist in Marl for anxious patients, tax advisor near me who speaks English, emergency plumber Recklinghausen Sunday. Keep the list fixed so months are comparable.
2. Ask each one in three places, in a logged-out window. ChatGPT, Perplexity and a Google search that produces an AI Overview. Logged out matters: your own account has months of history that will name you for reasons no stranger benefits from.
3. Record three things per question. Were you named, who was named instead, and which sources the answer cited. The third column is the useful one, because it is a list of the exact places you are missing from.
4. Change one thing, then wait a month. Not five things. Change one, so that when the answer moves you know which change moved it. This is slow and it is the only version of this work that produces knowledge rather than activity.

If you would rather not do it by hand, the free analysis runs the machine version of the same test. It reads your public pages, asks live answer engines the buyer questions for your trade, and shows you which sources they cite instead of you. It is the same procedure above, run in about three minutes, and you keep the report whether or not you ever speak to us. The AI Visibility Agent is what running it on a schedule looks like.

The honest part: what nobody controls

Nobody controls this, including the people charging monthly for it. There is no submission form, no ranking dial and no account manager at any of these engines who can add your name. There is only the evidence trail, and how consistent it is.

It moves in months, not weeks. Engines hold cached impressions of a business for a long time, and a firm that has been described inconsistently for five years does not become consistent to them in a fortnight. Expect the first movement one to two quarters after the housekeeping is genuinely done.

The engines also disagree with each other, so "get cited" is not one job. They draw on different source mixes, and a firm that is named readily in one may be absent from another for reasons that have nothing to do with quality. Test all three rather than optimising for whichever one you personally use.