AI visibility is not a single ranking
When a customer asks an AI search tool for a recommendation, the answer can vary with the platform, wording, location, time, personalisation and the live web information available to the system. That makes “What rank am I in AI?” the wrong first question.
The useful question is: for the commercially important questions our customers ask, how often is our business represented, cited or recommended, and what evidence is the AI using? Measuring that consistently gives you something much more valuable than a screenshot: a baseline, a pattern and a practical work list.
Google’s current guidance is clear that foundational SEO still matters in generative search. It recommends useful, original content, a clear technical structure and accurate business information rather than chasing shortcuts or creating pages for every possible query variation. Read Google’s official generative AI search guidance.
Start with evidence, not anxiety. A business that is not named in one response has found a research lead, not received a verdict. First establish whether the prompt was commercially relevant, what the answer cited, and whether the same pattern appears across a carefully selected query set.
Step 1: Define the questions that matter commercially
Do not build a list from every keyword tool variation. Start with the moments in which a customer is likely to compare providers, validate a decision or look for a trusted local answer. A Christchurch roofer might track emergency repair, re-roofing and material-choice questions. An Auckland accountant might track service, problem and comparison questions that reflect their actual client base.
| Query type | Example | Why it matters |
|---|---|---|
| Direct local service | “roof repair Christchurch” | Tests whether the business is present in a high-intent recommendation context. |
| Problem-led | “who can fix a leaking roof in Christchurch?” | Reflects conversational customer language and urgent needs. |
| Comparison / selection | “what should I look for in a Christchurch roofing company?” | Reveals which trust signals and sources the system surfaces. |
| Expertise question | “how long does a metal roof last in NZ?” | Tests whether educational content is useful enough to be selected as supporting evidence. |
For a small local business, a starting set of 10 to 15 carefully chosen queries is usually more useful than a list of 200 vague variations. Keep the query wording, target geography and intent in a simple document so you can repeat the test fairly.
Step 2: Separate the platforms and record the conditions
Do not treat every AI product as interchangeable. Google AI features draw on Google Search systems and can be measured in Search Console where the relevant reporting is available. Perplexity, ChatGPT and Gemini have their own interfaces, retrieval behaviour and citation patterns. A result on one platform should not be presented as proof of a result on another.
For each observation, capture the platform, query, date, country or location setting where applicable, whether your business was named, the cited URLs, named competitors and a short note on the wording of the response. A citation to your homepage, a service page and a third-party directory listing may represent very different opportunities.
Platform
Record the specific product tested: Google AI features, Perplexity, ChatGPT or Gemini. Keep their results separate.
Query and intent
Save the exact prompt, whether it is local, comparison-led, problem-led or informational, and the city if relevant.
Your presence
Note whether your brand, domain, page or listing was named or cited, rather than reducing everything to a vague score.
Evidence and competitors
Record cited URLs, competitors named, third-party sources and anything that explains why the answer was constructed that way.
Step 3: Build a baseline before changing everything
Run your defined query set once, document it and resist the temptation to react to every individual answer. The baseline is useful because it tells you where the gaps are. Perhaps AI recommends competitors for direct-service questions but cites your educational content for advice questions. Perhaps it finds your Google Business Profile but cannot connect it confidently to your service pages.
Use a simple tracking table with consistent fields. The point is not to create a beautiful dashboard on day one. The point is to retain the raw observations that explain change later.
| Field | What to record |
|---|---|
| Query | Exact text and search intent. |
| Platform and date | The product tested and when the result was recorded. |
| Business result | Named, cited by URL, mentioned without a link, or absent. |
| Evidence surfaced | Your pages, Google Business Profile, reviews, directories, articles or third-party publications. |
| Competitors and sources | The names and cited URLs that appear instead. |
| Action hypothesis | The next reasonable improvement to investigate, not an assumption of causation. |
Step 4: Use the findings to improve the right things
AI visibility improves for the same broad reasons that durable search visibility improves: clear crawlable pages, useful original information, accurate local business details, credible supporting evidence and a website that makes it easy to understand what the business does. Google explicitly warns that there is no special markup required for generative AI search and that structured data alone is not a shortcut. It remains useful when it accurately describes visible content and supports eligible search features. Google’s structured-data documentation explains this distinction.
- If AI does not understand the business: check entity consistency, service descriptions, local business details and the relationship between the brand, people and locations.
- If competitors own the relevant service evidence: improve the corresponding service page with actual customer questions, process information, pricing context where appropriate, proof and clearer internal links.
- If third-party sources dominate: improve genuine reviews, local citations and relevant external mentions rather than trying to manufacture links or mentions.
- If informational pages are cited but service pages are absent: strengthen the path from useful educational content to the related commercial page.
- If technical basics are weak: fix indexation, canonicals, page performance and structured data before writing another generic article.
Step 5: Re-test on a fixed rhythm
Once you have made a meaningful group of changes, re-run the same query set at a defined cadence. Weekly monitoring may make sense for an active AEO campaign; monthly may be enough for a small business with limited change volume. The important thing is consistency. A sudden movement in one platform can be useful to investigate, but longer-term patterns are the basis for a commercial decision.
Use the free AI Visibility Checker for a quick one-query starting point. If you need ongoing multi-platform monitoring for your own site or agency clients, CiteSpy was built by Arise SEO to support repeatable citation tracking. You can also work through our AEO Readiness Checklist to identify practical foundation gaps.
What not to measure
Do not claim a universal “AI rank” without being able to show the underlying queries and platforms. Do not assume a business is invisible because one unrepresentative prompt did not name it. Do not pay for inauthentic mentions. And do not publish thin pages solely because they contain a phrase you hope an AI model will repeat.
The measurable, sustainable objective is simpler: make it easy for people and search systems to find clear, credible information about your business at the moments that matter. That is good SEO, and it is the best foundation for durable visibility in AI-assisted search.