What AI systems need in order to describe your business correctly

Software describing your business is not reading it the way a customer does. It is not persuaded by your homepage headline and it does not notice that the photograph is of your own crew. It is looking for facts it can extract with confidence, and where it cannot find one it will either omit it or infer it from something nearby.

Most of the wrong descriptions business owners find are inference filling a gap. Understanding which gaps exist is most of the work.

The facts every system is looking for

There is a fairly stable set. If each of these is stated clearly, in text, and identically everywhere, you have removed most of the room for error:

• What kind of business this is, in a recognized category rather than a self-invented one.
• The exact legal or trading name, in one form.
• Where it is, and separately, where it works. These are different facts and are frequently conflated.
• What it sells, named the way customers name it rather than the way the industry does.
• When it is open, including the exceptions.
• How to reach it — a phone number, and whether that number takes texts.
• What it costs, at least in the sense of how pricing works.
• Whether it is currently trading. A business that has moved or rebranded and left the old records behind is a business that systems will keep describing in the past tense.

Structured data is how you say these things without ambiguity

Structured data — schema.org markup, usually written as JSON-LD in the page — is a standard vocabulary for stating facts in a form software does not have to interpret. Rather than hoping a system works out that "Open 7 til 5, closed Sundays" means what it means, you state opening hours in a defined format and there is nothing to work out.

For a local business the types that carry the most weight are LocalBusiness (or a more specific subtype where one fits your trade), Service for what you offer, and FAQPage for questions you answer. Adding the more specific subtype is worth the two minutes: "Plumber" says more than "LocalBusiness".

Where the contradictions usually come from

When a description comes back wrong, the cause is almost always one of a small number of things:

1. An old listing nobody remembers creating, with a previous phone number or address, still ranking well enough to be read.
2. A former business name still present on directories, or a name that appears with and without a suffix — "and Sons", "LLC", "Inc" — depending on the source.
3. A service that was dropped years ago and is still described on a page nobody links to any more.
4. Hours that differ between the website and the Business Profile, usually because one was updated and the other was not.
5. Two phone numbers in circulation, one of which was a tracking number from a campaign that ended.
6. A service area described one way on the site and another way on the profile.

None of these are exotic, and all of them are findable in an afternoon by searching your own business name and reading what comes back with the assumption that it is wrong until confirmed.

Text beats pictures, every time

A surprising amount of important information on small business sites exists only inside images: hours in a graphic, a price list as a photographed sheet, a service menu as a scanned PDF, contact details baked into a banner.

This information is effectively invisible. It is also invisible to a customer using a screen reader, which is a separate and more important problem. The fix is to have the text as text, with the image as decoration alongside it if you want the image.

Answer the questions you already get asked

The most useful content you can add is the set of questions you answer on the phone every week. You already know them: do you cover this area, how soon can you come, do you charge for an estimate, do you work weekends, do you handle the permit, how long will it take.

Written down, in your own words, these do three things at once. They answer the visitor who arrived with exactly that question, they give search engines something specific to match, and they give an AI system something to cite rather than infer. Marked up as an FAQ, they are unambiguous.

Write them the way you would say them. Vague, hedged answers are worse than no answer, because they occupy the space where a real answer would go.

A one-afternoon audit you can do yourself

1. Search your business name. Read every result on the first two pages and note anything wrong or out of date.
2. Ask two AI assistants about your business by name, and then about your category in your area. Note what they get wrong and what they omit.
3. Open your own site on a phone and check whether hours, service area, phone number and services are readable as text.
4. Compare your website details against your Business Profile, field by field, and make them identical.
5. Write down the six questions you get asked most and answer them in writing.
6. Check whether your site has structured data at all. Google’s Rich Results Test will tell you free.