How to check whether AI search can find your business
A practical AI search visibility check for Australian small businesses: test real customer questions, verify website signals and measure results without chasing unreliable AI rankings.
A Melbourne business owner asks ChatGPT for a local supplier. Another customer asks Google which service suits their situation. Neither is searching for the business by name.
Will your business be easy to discover, understand and compare?
That is a useful question. But opening an AI assistant once, asking it to recommend your business and treating the answer as a ranking report is not a reliable test. AI answers vary with the wording, location, available sources and product experience. A business can be mentioned without receiving a click, or receive enquiries from people who researched it across several channels.
A more useful AI search visibility check looks at three separate things: whether your website can be discovered, whether it clearly explains the business, and whether customers can reach a relevant next step.
Start with five questions your customers actually ask
Choose questions that reflect a buying decision, not just your brand name. For a hypothetical Melbourne commercial cleaning company, they might be:
- Who provides after-hours office cleaning in Melbourne?
- What should I ask before choosing a commercial cleaning provider?
- Can a cleaning company handle multiple office locations?
- How do commercial cleaning contracts usually work?
- What information should I prepare to request a quote?
These are illustrative questions, not measured keyword volumes.
Notice the difference between them. The first asks for providers. The second and fourth ask for guidance. The third tests a specific capability. The fifth tests the path to an enquiry. A business website should not try to turn every question into a sales pitch, but it should give useful, accurate answers where it has expertise.
Write down the question, intended customer, location and the page on your website that ought to answer it. If you cannot identify that page, you have found a content or architecture gap before opening an AI tool.
Test AI answers as observations, not permanent rankings
Run the same small set of questions in the AI search experiences your customers plausibly use. Record the date, product, prompt, location context where known, whether your business is mentioned, whether a source link is provided and whether the information is accurate.
Use a simple worksheet:
| Question | Platform and date | Mentioned? | Source linked? | Correct information? | Follow-up |
|---|---|---|---|---|---|
| After-hours office cleaning in Melbourne | Record your actual test | Yes / No | Yes / No | Yes / No | Check service and location page |
| What to ask a cleaning provider | Record your actual test | Yes / No | Yes / No | Yes / No | Improve buyer guidance |
Leave cells blank until you have performed the test. Do not turn one result into a claim such as “ranked number three in ChatGPT”. Repeat important prompts over time using consistent conditions, but expect variation. If you are comparing providers, use the same questions and observation rules for each; do not treat the resulting counts as market share.
A missing mention is a diagnostic starting point, not proof that an AI platform has penalised your website. It may reflect the question, available evidence, relevance or the assistant's choice of sources.
Check the pages behind the answers
Open the service page a customer should land on. In the first screen and main body, can someone answer these questions?
- What does the business actually provide?
- Who is the service for, and where is it available?
- What is included, and what is outside scope?
- What evidence helps a buyer trust the claims?
- What should a visitor do next?
For the cleaning example, a vague page saying “premium solutions for every business” is less useful than a clear explanation of service areas, after-hours availability, types of premises, quotation requirements and limitations. Do not invent service areas, certifications, client logos or review scores to fill gaps.
Then check the relationships. Can a visitor reach the relevant service page from navigation? Does a genuinely useful guide link to it? Is the contact path obvious? If several articles target the same buyer question with near-identical answers, improve one authoritative page rather than publishing another variation.
This is the practical work behind AI Search Optimisation: clarify the business and the path through its content, not simply add more mentions of “ChatGPT” to every page.
Check technical eligibility before buying an AI SEO tool
Google's May 2026 guidance for generative AI search says conventional SEO foundations still matter. Important pages should be accessible, indexable and useful. It also says Google does not require a special AI text file, unusual content chunking or AI-specific schema markup for its generative search features.
A focused technical review should confirm:
- Important public service pages return a successful response and are not accidentally blocked by
noindexor access controls. - Canonical URLs identify the intended pages, and the sitemap reflects the published content.
- The main answer is available in readable page content rather than hidden behind an interaction that a crawler may not complete.
- Titles, headings and internal links help people understand the page's purpose.
- Structured data, where present, describes the visible business and content accurately.
Google's structured data policies explicitly warn that valid markup does not guarantee a rich result. Adding unsupported review ratings or an elaborate schema graph is not a shortcut to an AI recommendation.
ChatGPT search uses a different crawler arrangement. OpenAI's crawler documentation distinguishes OAI-SearchBot for search from GPTBot for model training. The preferences are independent. If ChatGPT search visibility matters, review whether search access is permitted and whether firewall rules contradict that preference. Do not assume that allowing a bot guarantees inclusion.
For a deeper explanation of crawler categories, edge controls and training preferences, see how to manage AI crawler access without losing discoverability. This check is about overall discoverability; crawler policy is only one part of it.
Verify business facts beyond your own website
Customers may encounter your business through maps, directories, partner websites, professional profiles and other third-party references. Check whether your name, website, location or service area and core offering are accurate and consistent wherever you control those details.
A service business operating across Australia should not imply it has a staffed office in every city. A mobile provider should describe its real coverage rather than creating misleading location pages. A portfolio example should make clear what work was actually delivered.
Useful independent evidence might include a genuine case study, an accurate professional profile or a relevant industry listing. Buying bulk directory placements or manufacturing reviews does not create trustworthy evidence.
Measure outcomes without inventing an AI visibility score
Use measurement that answers different questions:
| Signal | What it can tell you | What it cannot prove |
|---|---|---|
| Search Console query and page data | Google impressions, clicks, CTR and position for available reports | Your ranking in every AI assistant |
| Search Console generative AI reporting, where available | Visibility in Google's reported generative experiences | Complete visibility across all platforms |
| Analytics referral sources | Visits attributed to identifiable AI or search referrals | Every AI-influenced enquiry |
| Recorded prompt observations | Whether specific answers mentioned or linked to you at a point in time | A stable or universal recommendation rank |
| Qualified enquiries | Whether visitors contact you about the right services | That one AI answer caused the conversion |
Google's official generative AI optimisation guide describes a dedicated generative AI performance report in Search Console. Use it if your property has access, but do not claim the data exists before checking. Different platforms expose different levels of reporting.
Ask new enquiries how they found you, without forcing them into a single channel. Someone may first see an AI answer, verify your work in Google, then contact you directly. Treat attribution as useful but incomplete evidence.
A practical 30-day improvement plan
Week one: establish a baseline. Pick five to ten real customer questions. Record current answers and links. Check Search Console and analytics access. Identify the two or three most commercially important pages.
Week two: fix clarity. Improve those pages with specific service scope, eligibility, locations, evidence and next steps. Correct contradictory business details. Add contextual internal links only where readers genuinely benefit.
Week three: fix access. Review indexing, canonical URLs, sitemap entries, structured data accuracy and crawler policies. Test the real rendered pages, not only a configuration screen. If a technical change could affect existing search traffic, document it and keep a rollback path.
Week four: retest and review. Repeat the recorded prompts under comparable conditions. Check whether the information is more accurate and the next step clearer. Review search and referral data where available. Do not declare success from a single new mention.
If the baseline reveals a serious indexing problem, fix that before spending the month producing new articles. If the site is technically sound but the offer is unclear, content structure may be the better investment.
What not to buy first
Be cautious about a service that promises a fixed ChatGPT ranking, guaranteed citations, or a proprietary “AI visibility score” without showing how it is calculated.
Also be cautious about a package whose main deliverables are llms.txt, generic FAQ pages and large volumes of AI-written content. Those may have specific uses, but they do not replace accurate services, evidence, crawlability and helpful information. Google's guidance explicitly says llms.txt is not needed for its generative search features.
The most useful first engagement may be a focused audit and a handful of corrections, not a monthly content subscription.
The next decision
The aim is not to persuade every AI assistant to say your business name. It is to make your business easier to find, accurately describe and confidently choose when a relevant customer is researching a problem.
Start with the questions, verify the pages, check the technical foundation and measure the enquiries that matter. Then invest in the gaps you can actually demonstrate.
If you want an independent review of how your business is represented across search and AI-assisted discovery, start a conversation about your website. A focused assessment can separate urgent technical issues from content improvements that are worth making over time.