Guide
How to get recommended by ChatGPT
Why ChatGPT names some businesses and not others
When a customer asks "who should I hire for X in Y?", ChatGPT doesn't search a ranked index the way Google does. It assembles an answer from two places: what its models learned in training, and what live retrieval can read right now. In both cases the same rule applies: it names businesses whose existence, services and reputation it can read and cross-check. If your site blocks AI crawlers, buries what you do in images and marketing copy, or your details conflict from one directory to the next, the model recommends someone it's more confident about.
Everything below serves that one rule. This is the same playbook we run on our own network of 70+ Australian sites, including this one.
The seven steps, in order
1. Let the AI crawlers in
Your robots.txt decides whether OpenAI's crawlers can read you at all. Allow GPTBot (training), OAI-SearchBot (search index) and ChatGPT-User (live visits when a user's question triggers browsing). If you use Cloudflare, check its AI-bot blocking isn't overriding your robots.txt: Cloudflare ships settings that serve a block-everything file without ever touching your origin copy. Many Australian sites are invisible to ChatGPT for this reason alone.
2. State what you are in plain text
Add an llms.txt file at your site root: a short plain-text summary of what the business does, where it operates, what it sells and how to make contact. Then check your about and service pages say the same things in real sentences, not just imagery. An AI can only repeat facts someone wrote down.
3. Give it structured data to quote
JSON-LD schema (Organization, LocalBusiness, Service, Product, FAQPage) is the difference between the model inferring what you do and knowing it. Mark up your name, address, phone, services, prices where you publish them, and FAQs. Test that it parses; broken schema is worse than none.
4. Build answer-shaped pages
One page per question, answer first. "How much does a coolroom cost in Sydney?" deserves a page that opens with the honest range, then the detail. Humans skim past pages that open with three paragraphs of brand story, and so do models. Headings that match real questions, direct answers under each, specifics over slogans.
5. Get corroborated
Models trust claims other sources repeat. For Australian businesses that means: a claimed, current Google Business Profile; reviews that mention specific jobs, suburbs and outcomes (a review that says "installed a 40m² coolroom in Penrith in March" is worth ten that say "great service"); listings in the directories that matter for your trade; and identical business name, address and phone everywhere. Conflicting details read as risk, and models are built to avoid risky recommendations.
6. Own something specific
"Plumber" is a coin-toss. "Backflow testing in the Illawarra" is a recommendation a model can make with confidence. Pick the intersections of service and place you actually want, and give each one a real page with real proof. Specific beats big, and for local questions the competition is thinner than the head terms.
7. Measure with transcripts, not vibes
Ask ChatGPT the questions your customers ask, the same set every month, and keep the transcripts. That's the scoreboard: which questions name you, which name competitors, and whether this month's fixes moved anything. It's how we report to clients, because nobody can promise a specific answer and the transcript doesn't lie.
What doesn't work
- Buying the organic answer. Nobody sells it. Labelled ChatGPT ads are live in Australia, but they're a separate, marked placement, not the recommendation itself.
- Guarantees. "We'll get you recommended in 30 days" is a red flag, not a service. Movement is measurable; specific answers aren't promisable.
- Stuffing pages with "best in Australia" claims. Unverifiable superlatives don't corroborate. Specifics do.
Where to start
Run the ten-minute self-test: ask ChatGPT three questions your customers would ask, and see who it names. If you're not there, our free AEO audit does the full version: real transcripts of what AI says about you, a technical score on the signals above, and the three fixes that matter most, within one business day. ChatGPT-specific version with the free visibility check: chatgpt-ads.com.au.
The short answers
How does ChatGPT decide which businesses to recommend?
From what it can read and verify. When someone asks "who should I hire?", ChatGPT assembles a shortlist from its training data and live retrieval: websites its crawlers can access, facts stated in plain machine-readable form, and corroboration from reviews, directories and other sites. Businesses it can't read, or whose details conflict across the web, don't make the shortlist. No payment is involved in the organic answer.
How long does it take to show up in ChatGPT's answers?
It varies by answer type, so be wary of anyone quoting a fixed number. Answers that use live web retrieval can reflect site fixes within weeks; answers drawn from model memory alone move slower, on model-update timescales. That's why we measure with monthly transcripts of the actual questions customers ask, rather than promising a date.
Can I just pay to be recommended by ChatGPT?
No. The organic recommendation is not for sale from OpenAI or anyone else. Paid ads inside ChatGPT do exist for Australian advertisers via OpenAI's Ads Manager, but they're labelled ads, separate from the organic answer. The organic spot is earned through readability, structure and corroboration.
Does this work for small local Australian businesses?
Yes, and often faster than for big brands, because local questions have thinner competition. A local business with an up-to-date Google Business Profile, reviews that mention specific jobs and suburbs, consistent details everywhere, and a website that states plainly what it does and where, is exactly what an AI can confidently recommend.