When someone asks an assistant which business to hire, the answer names two or three and everyone else does not exist. There is no page two to be on. This is a guide to what decides which businesses get named — and, just as usefully, what nobody can promise you.
Check you are not blocking the crawlers
Start here, because it is free and a surprising number of sites fail it. AI assistants use their own crawlers, and plenty of sites block them by default, or by way of a security product nobody configured deliberately. If GPTBot cannot read your site, no amount of content work will help.
| Crawler | Who runs it | What it feeds |
|---|---|---|
| GPTBot | OpenAI | ChatGPT training and retrieval |
| OAI-SearchBot | OpenAI | ChatGPT search results |
| ClaudeBot | Anthropic | Claude |
| PerplexityBot | Perplexity | Perplexity answers and citations |
| Google-Extended | Gemini and AI Overviews |
Check your robots.txt for these. Allowing them is a decision worth making deliberately either way — for a services business, being quotable is a referral channel.
Make your business resolve as one thing
Models have to work out that the business on your website, the one in the directory listing, and the one on the review site are all the same organisation. When those records disagree — different address formats, a business name with and without the legal suffix, two phone numbers — the model either picks one at random or hedges by not naming you.
The fix is consistency and structured data: one canonical name, address and contact across every surface, and schema markup on your site that describes the business once rather than redefining it on every page. Worth knowing: self-serving review markup on your own business does the opposite of helping, because it violates search engine policy and can get the whole markup discounted.
Publish things that are quotable
This is where most of the gain is, and it is the part agencies tend to skip because it requires saying something specific. Models quote numbers, ranges, comparisons and direct answers. They skip positioning statements. A page saying you offer competitive pricing and exceptional service contains nothing a model can lift into an answer, so it does not get lifted.
- ✓Real price ranges rather than a contact us for pricing page
- ✓Direct answers stated in the first line, before the context and caveats
- ✓Comparison tables — the single most quoted format in AI answers
- ✓Specific claims a model can attribute to you and repeat safely
The test that tells you where you stand
Ask ChatGPT, Perplexity and Google AI the question a customer would ask: who should I hire for [your service] in [your town]. Then ask what they know about your business by name. Most owners doing this for the first time find either nothing, or something confidently wrong. That result is your starting baseline, and it is free to check.
What cannot be promised
A position, because there is not one. Answers vary between users, between sessions and between model versions. Anyone quoting you a guaranteed spot in an AI answer is describing something that does not exist as a product.
What can be measured is citation rate for a defined set of buying questions, tested on a schedule across the major assistants; referral traffic from assistant domains, which shows up in analytics; and the conventional signals that move alongside, since this work usually improves featured snippets and AI Overview appearances too.
Where it fits in the budget
Alongside SEO rather than instead of it. The underlying work overlaps enough that treating them as separate retainers means paying twice for the same technical foundation. If your SEO is already strong, this is an add-on measured in weeks. If it is not, fix that first — the same work serves both.