How to Show Up in ChatGPT, Claude and AI Search

Somebody looking for a plumber, an accountant or a supplier increasingly asks an assistant rather than a search engine. They get a short answer naming two or three businesses, and the ones not named never find out they were considered.

This has produced an industry of people selling "AEO", "GEO" and "AI optimization", much of which is ordinary SEO with a new invoice attached. Some of it is real. Here is the difference.

How an Assistant Decides Who to Name

Broadly, three routes, and they need different things from you:

• Training data. What the model absorbed long ago. You cannot influence this on any useful timescale, and anyone selling you "training data optimization" is selling you a feeling.
• Live retrieval. The assistant searches the web during the conversation and reads what it finds. This is the one you can affect, and it behaves a lot like classic search with a much lower tolerance for ambiguity.
• Third-party sources. Directories, review sites, industry listings, news. Assistants lean heavily on these for local and commercial questions because they are structured and comparatively trustworthy.

What Is Actually Worth Doing

1. Be Readable Without JavaScript

This is the single biggest technical differentiator. Many retrieval fetchers do not execute JavaScript, so a page whose content is assembled in the browser can arrive as an empty shell. Test it: disable JavaScript, load your key pages, and see what is left. If your services and location vanish, that is what a large share of automated readers see.

2. Answer the Question in the First Two Sentences

Summarizers reward passages that state the answer plainly and immediately. A services page that opens with "For over a decade, our passionate team..." gives a retriever nothing to quote. One that opens with "We install and repair commercial refrigeration in Perth, usually same-day" gives it a sentence it can lift whole.

3. Make the Facts Unambiguous and Consistent

Name, address, phone, hours, service area, what you do and do not do — identical everywhere they appear, on your site, your Google Business Profile and every directory. Assistants weigh corroboration heavily; a business whose hours disagree across three sources is one an assistant will hedge about or skip.

4. Use Structured Data Properly

Organization, LocalBusiness, Service, FAQPage and Product markup are machine-legible statements of the facts you just made consistent. They are not a ranking trick; they remove the need for anything to infer. Mark up what is genuinely on the page and nothing that is not.

5. Publish an llms.txt, but Understand What It Is

A plain-text file at the root summarizing what your business does, what things cost and where the important pages are. It is a convention rather than a standard, and no assistant is obliged to read it. It costs an hour, it makes a machine-readable summary of your business exist, and it is trivially easy to let it go stale — which is worse than not having one. Ours is generated from the same configuration the site renders from, so it cannot drift: you can read it at multitasker.llc/llms.txt.

6. Decide Deliberately Whether to Allow AI Crawlers

GPTBot, ClaudeBot, PerplexityBot and others can be allowed or blocked in robots.txt. There is a real trade-off and it is yours to make: blocking protects your content from being absorbed, and also removes you from answers that would have named you. For most small businesses selling a local service, being findable is worth more than the content is. For a publisher whose archive is the product, the sum is different.

7. Get Named Somewhere Other Than Your Own Site

Assistants corroborate. Industry directories, local chambers, supplier listings, genuine press. Not bought links — those have always been a bad trade and the models are not fooled by a private blog network any more than Google is.

What to Ignore

• "Guaranteed ChatGPT ranking." There is no ranking to buy. There is no placement, no ad slot, and nobody at OpenAI or Anthropic taking your money for one.
• Keyword stuffing for models. The thing you are optimizing for is a language model. It is exceptionally good at noticing writing bent out of shape.
• Hidden text aimed at crawlers. Cloaking is against every search engine's guidelines, it is detectable, and the downside is losing the traffic you already have.
• Rewriting your whole site as FAQs. A page that is genuinely a list of questions should be one. Most pages are not.

How to Know If It Is Working

This is the honest part: measurement here is poor. Assistants do not send referrer data the way a search engine does, so attribution is patchy at best.

What you can do is ask. Write down the ten questions a customer would actually type — "best commercial electrician in Fremantle", "who does emergency refrigeration repair near me" — and put them to the assistants with search enabled, on a schedule, recording whether you are named and who is named instead. It is a manual, honest measurement, and it beats a dashboard number nobody can explain.

That is exactly what our own Findable does automatically: it asks your buyers' questions with search on and records whether you appear, alongside checking whether your content survives without JavaScript, whether your structured data is valid and whether AI crawlers can reach you. The free check covers the technical half with no signup.

The Uncomfortable Conclusion

Most of "AI SEO" is the technical hygiene people were already meant to be doing: crawlable pages, plain answers, consistent facts, valid markup, real corroboration. The assistants simply punish ambiguity harder than a search engine does, because they have to produce one answer rather than ten links.

If your site is slow, thin or assembled in the browser, fixing that will do more for your visibility in AI answers than anything sold to you with "AEO" in the name. Start with why your site is slow and the local SEO checklist — the foundations are the work.

Common Questions

How Do I Get My Business Mentioned by ChatGPT or Claude?

Mostly through live retrieval, which behaves like search: be crawlable, make your content readable without JavaScript, state what you do and where in plain text near the top of the page, keep your name, address, hours and service area identical everywhere, use valid structured data, and get listed in credible third-party directories. You cannot influence what a model was trained on.

Is AEO or GEO Different from SEO?

Largely not. The technical fundamentals are the same, and most of what is sold as AI optimization is ordinary SEO hygiene. The genuine differences are that many retrieval fetchers do not run JavaScript, and that assistants punish ambiguous or inconsistent facts harder than a search engine because they must produce a single answer rather than a list.

Should I Block AI Crawlers in robots.txt?

It is a real trade-off rather than a best practice. Blocking GPTBot, ClaudeBot and similar protects your content from being absorbed, and also removes you from answers that would have named you. For a local service business, being findable is usually worth more than the content is. For a publisher whose archive is the product, the calculation is different.

What Is llms.txt and Do I Need One?

A plain-text file at your domain root summarizing what your business does, what it costs and where the key pages are. It is a convention, not a standard, and no assistant is required to read it. It takes about an hour and is easy to let go stale — a stale one is worse than none, so generate it from the same source your site uses if you can.

Can I Track Traffic from AI Assistants?

Only partially. Assistants generally do not pass referrer data the way search engines do, so attribution is unreliable. The practical alternative is to write down the ten questions a real customer would ask, put them to the assistants with search enabled on a regular schedule, and record whether you are named and who is named instead.