To improve the chance that ChatGPT recommends your business, keep your site accessible to search crawlers, publish the clearest answer to the buyer’s exact question, support it with current facts a person can verify, earn relevant independent coverage, and recheck the same questions over time. None of those steps guarantees inclusion. They create stronger evidence for retrieval and selection.
1. Make the business eligible to be found
OpenAI says any public website can appear in ChatGPT search. If you want page content to be used in summaries and snippets, do not block OAI-SearchBot. That bot is for search discovery; it is separate from GPTBot, which controls potential model-training access.
- Keep important service, product, location and policy information in visible page text.
- Give each useful topic a crawlable URL and link to it from relevant pages.
- Use one descriptive page title, one clear main heading and an accurate summary.
- Check robots rules, canonical URLs and index status before rewriting the copy.
This is the same technical foundation Google recommends for conventional and AI-assisted search. Special “AI files” are not a substitute for an accessible, useful website.
2. How does ChatGPT find a business to recommend?
ChatGPT may search the web when a question needs current information. OpenAI says search can rewrite a person’s question into more specific searches, use approximate location for local results, and return citations to relevant sources. Search results are ranked using multiple factors intended to surface relevant, reliable information, but OpenAI does not publish a simple formula and does not guarantee placement.
That creates two separate hurdles. First, the business must be retrievable: its important pages and facts need to be accessible. Second, it must be a defensible answer to the specific request. Being indexed clears the first hurdle; it does not automatically clear the second.
3. Answer the question a customer actually asks
A generic About page rarely answers a recommendation question. Build pages around real decisions: who the business is for, what problem it solves, where it operates, what it costs, and when it is or is not the right choice.
Use the customer’s full situation
“Best accountant” is vague. “Accountant for a Sydney construction business using Xero” has a service, market and constraint. Your page should make those real boundaries explicit rather than repeating a broad keyword.
Lead with the direct answer
Put the useful conclusion before the company history. Then support it with the services, products, process, prices or price conditions, service area, limitations and next step a buyer needs.
4. Publish facts that can be checked elsewhere
OpenAI’s shopping documentation says relevance, availability, price, quality and seller identity can contribute to product and merchant selection. A service business has different facts, but the principle is the same: make the decision-critical details current, specific and consistent.
| Weak evidence | Useful evidence |
|---|---|
| We are the best choice | Named use case, real constraint and why the fit is strong |
| Affordable pricing | Current price or a clear explanation of what changes the quote |
| Trusted by customers | Genuine reviews, case evidence and an identifiable source |
| Serving everywhere | Actual cities, regions or delivery areas and their limits |
Independent reviews, professional directories, editorial coverage and community discussions may provide corroborating evidence. Earn that coverage where customers already research the category; do not manufacture mentions or buy generic links.
5. Why does ChatGPT recommend a competitor instead?
A competitor appearing is not proof that it has a secret AI optimization trick. It may fit the question more precisely, expose clearer facts, have stronger independent corroboration, or simply appear in the sources retrieved for that answer. Diagnose the observed answer before changing pages.
| What you observe | What to inspect next |
|---|---|
| ChatGPT cannot find the business | Crawler access, index status, important page text and internal links |
| It describes the wrong service or location | Conflicting facts across the website, profiles and directories |
| A competitor is named instead | The exact fit, proof and cited sources behind that answer |
| The business appears once, then disappears | Answer variability — repeat the same question and context |
6. Can you pay or submit a business for recommendation?
There is no general submission that buys an organic ChatGPT recommendation. Where ChatGPT ads are available, a business can pay for a clearly labelled ad placement. OpenAI says ads remain separate from answers and do not influence the organic answer. Product merchants may also have separate feed or commerce onboarding routes, but those do not guarantee that a product or business will be selected.
7. Skip the shortcuts
- Keyword stuffing: repeating “best business” does not create proof.
- Mass-produced near-duplicate pages: one useful page should own one intent.
- Schema as a recommendation guarantee: supported markup can clarify visible facts, but it cannot force selection.
- llms.txt as a ranking trick: it is not a Google ranking or AI-citation shortcut.
- One-off screenshots: a single answer is an observation, not a permanent rank.
8. Measure the exact question, not a made-up score
Record the question, market, engine, date, answer, named businesses and cited sources. Repeat the same useful questions because AI answers and retrieval can vary. Watch whether the business begins appearing more often and whether real referral visits or enquiries follow.
Citano’s free quick check uses one live OpenAI model to create a directional observation. It does not reproduce every consumer ChatGPT surface, location or personalized conversation. The report keeps the question and answer visible so you can inspect the evidence rather than trust a mystery score.
- 01Question
Save the exact buyer wording. Do not paraphrase it between tests.
- 02Context
Record the product, market, location, date and whether web search was used.
- 03Answer
Capture named businesses, their order, the answer text and every cited source.
- 04Outcome
Track referral visits, enquiries or sales separately from answer visibility.
9. How long does it take to improve?
There is no fixed waiting period. You can verify a technical fix immediately, but recrawling, retrieval and answer selection are controlled by the search systems, not the site owner. Use a dated baseline, improve one high-value page, then repeat the same tests on a regular schedule. A movement in repeated observations is evidence; a promised number of days is not.
A practical 30-day starting plan
- Day 1: record five real buyer questions and the businesses ChatGPT names.
- Days 2–5: confirm the important pages are public, crawlable, canonical and internally linked.
- Week 2: improve the one page that best answers the highest-value question.
- Week 3: correct inconsistent facts and work toward one relevant, authentic independent reference.
- Week 4: rerun the same questions and log the answers, citations and referral traffic.
Start with one observed answer
Check a public business website, inspect the buyer question and see which evidence appears before deciding what to change.
Check my websiteKeep reading
- What is Answer Engine Optimization? — how AEO overlaps with SEO.
- Citano methodology — what one observation can and cannot prove.
- Shopify AI visibility guide — the ecommerce-specific version.
Sources
- OpenAI, Publishers and developers FAQ — website discovery and OAI-SearchBot
- OpenAI, Searching the web with ChatGPT — search, citations, location and ranking limits
- OpenAI, Our approach to advertising and expanding access to ChatGPT — separation between ads and organic answers
- OpenAI, Shopping with ChatGPT Search — product and merchant selection factors
- Google Search Central, AI features and your website — crawlability, useful content and technical foundations
- Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024) — controlled benchmark of content interventions