# How does a local business get recommended by AI?

When someone asks an assistant who fixes roofs near them, it searches the web and names businesses whose own pages say plainly what they do and where. A local business gets there by saying both in words and in its data, giving a phone that dials, and linking the profiles that vouch for it, with nothing it cannot back.

By clickd. Published October 27, 2026. Updated November 1, 2026. About 3 minutes to read.

At a glance:

- 14%: Of trades pages have no tap-to-call link. Over the 2,438,096 trades pages it applies to, as of November 1, 2026.
- 7.0%: Of trades pages give no way to get in touch. Over the 2,529,908 pages it applies to.
- 20%: Of trades pages carry no structured data that parses. Over the 2,529,908 pages it applies to.
- 21%: Of trades sites have pages that nearly copy each other. Over the 57,230 trades sites read.

## What does an assistant need from a local business?

The same things a person needs, said where a machine reads them. The service and the town in the title and the opening line. An address or the area it serves, in words and in structured data. A phone number that dials when tapped. Its Google Business Profile and other profiles named in its data and linked from its pages, so the business on the page and the business on the map are plainly one.

On trades sites that serve a place, 14% of the pages clickd has read have no tap-to-call link, and 7.0% of trades pages give no clear way to get in touch: no form, and no link to ask for a quote or a booking.

## How do trades sites do?

clickd places each site it reads in a line of work. These are the rules in its rulebook that bear on a local business, trades sites against every site.

Pages that pass, trades sites against every site:

| What clickd checks | Trades pages passing, higher is better | Every site's pages passing | Trades pages counted |
| --- | ---: | ---: | ---: |
| A tap-to-call link (every page of a site that serves a place) | 86% | 39% | 2,438,096 |
| A way to get in touch (every page) | 93% | 83% | 2,529,908 |
| Structured data that parses (every page, JSON-LD as a search engine reads it) | 80% | 63% | 2,529,908 |
| Data of the right type (home, article and service pages) | 89% | 84% | 1,232,220 |
| The words in the HTML (home, article and service pages, before any script runs) | 96% | 89% | 1,232,220 |
| No sales claims it can't back (every page) | 96% | 97% | 2,529,908 |
| Ratings only with reviews shown (every page) | 95% | 97% | 2,529,908 |

_Counted over the 2,667,867 pages clickd has read on the 57,230 sites it has placed among the trades, each site at its latest read, as of November 1, 2026, against every site it has read._

| What clickd checks of the whole site | Trades sites passing | Every site passing |
| --- | --- | --- |
| Profiles named in its structured data | 65% | 81% |
| Profiles linked from its pages | 60% | 56% |
| Pages that are not near-copies of each other | 79% | 82% |

_Counted over the 57,230 sites clickd has placed among the trades, each at its latest read, as of November 1, 2026, against the 316,711 sites it has read in all._

## What about a page for every town?

A page for each town a business serves is worth having when each one says something true of that town: the jobs done there, how far the crew travels, what is different about it. A page that only swaps the town's name into the same text is what search engines call scaled content. 21% of trades sites already have pairs of pages that share half or more of their text.

## What does not work?

- Reviews and ratings in structured data that the page does not show. Search engines treat them as claims, and they can cost a site its rich results.
- Superlatives with nothing behind them, such as top rated or the area's favorite. They read as advertising to a person and as noise to a machine.
- A different name, address or phone number on the site, the profile and the directories. Make them say the same thing, everywhere.
- A profile on the map with no site behind it that says what the business does. The profile is found; the site is what an answer can quote.

## How do you fix it?

Read the site the way a crawler does, fix what is worth the most first, and read it again once the change is live. See [how to get recommended by ChatGPT](https://clickd.com/guides/get-recommended-by-chatgpt) for the crawler side, and [agentic SEO](https://clickd.com/guides/agentic-seo) for handing the fixes to an agent.

Make a local business readable to AI. Hand this to your agent:

> Use clickd to read yoursite.com. Make sure every page names the service and the town, every phone number is a tap-to-call link, and the structured data says what the business is, where it works and which profiles are ours, using only facts the pages already state. Fix the three things worth the most, one at a time, then have clickd read the site again after I deploy.

The calls it makes: `mttr_brief({ site: "yoursite.com" })`, `mttr_rescan({ site: "yoursite.com" })`, `mttr_changes({ site: "yoursite.com" })`.

A score or a rule passed says pages are well formed. It does not say an assistant cites the site. A site an assistant can read is one it can recommend; whether it does is the assistant's to decide.

## Sources

- [Local business structured data](https://developers.google.com/search/docs/appearance/structured-data/local-business): Google Search Central, on LocalBusiness data
- [Guidelines for representing your business on Google](https://support.google.com/business/answer/3038177): Google Business Profile help
- [Spam policies for Google web search](https://developers.google.com/search/docs/essentials/spam-policies): Google Search Central, including scaled content abuse

## How the figures were counted

Every figure here is counted from the 316,711 sites clickd has read, each at its latest read, with drafts left out, as of November 1, 2026. A share of pages counts only pages that were read. A score or a rule passed says pages are well formed. It does not say an assistant cites the site.