AI search
How to measure AI search visibility without guessing
Choose a fixed set of customer questions, ask each search-enabled engine, and record the date, language, location settings, answer and linked sources. Separate a brand mention from a citation of your website. Repeat selected questions and compare monthly under similar conditions. This measures the questions you sampled, not your visibility across all AI answers.
Why this exists
A lot of what is sold as “AI search optimization” is unmeasured. Schema gets added, an FAQ block appears, a file called llms.txt gets published, and the invoice says “AI optimization”, with nothing checking whether any of it changed anything.
This method is designed to make the observations inspectable, including questions where a business does not appear. A mention, a linked citation, a referred visit and a qualified enquiry are different outcomes; report each separately.
The method
1. The question set
Start with 12–20 questions across four categories, fixed so results are comparable over time:
| Category | Example |
|---|---|
| Direct brand | “What is Novarius?” |
| Service + qualifier | “Who builds websites for contractors?” |
| Problem-led | “Why isn’t my contractor website getting calls?” |
| Local | “Web design companies in Miami” |
The question set does not change between runs. Changing the questions makes results incomparable. That is the mistake that turns this from data into anecdote.
2. The systems
Google AI Overviews, Google AI Mode, ChatGPT Search, Perplexity, and Bing Copilot.
3. What gets recorded
For each question and each system: the full answer text, every source cited in order, whether novarius.us appeared and where, whether we were cited, mentioned without citation, or absent, and the date.
4. Controls
Use a fresh conversation for each question. Record engine and mode, whether search is enabled, language, date, signed-in status and available location settings. Logging out does not remove IP-based location or all personalization. Mark settings you cannot control as unknown. Save the answer and cited URLs, with screenshots or exports where available.
How to interpret the observations
Repeat selected questions to observe variation. Calculate citation rates only over completed observations, with the denominator shown. Untested questions are not failures.
Separate brand from discovery. A result for “What is Novarius?” does not demonstrate that the site appears when someone asks who builds contractor websites.
Record when no AI answer appears. In Google, that outcome is different from an AI answer that cites other sites. In other systems, distinguish answers with retrieval from answers without it.
Connect citations to business outcomes. Check referred visits and confirmed enquiries separately. See our conversion tracking service for the distinction between contact clicks and confirmed leads.
Do not assume causation. A change in citations can coincide with a page update while models, competitors and source selection also change. Keep the question set and conditions as consistent as possible.
The part that makes this hard
Chen and colleagues’ 2025 study observed a preference for third-party sources in its sample. Its results also varied by engine and language. That does not establish a universal selection rule or guarantee that a directory listing will produce a citation.
Google’s guidance for AI features says no special AI files or schema are required for AI Overviews or AI Mode. Start with accessible pages, clear information and markup that matches visible content.
What appears to move it: content genuinely worth citing, a consistent business identity everywhere it appears (the same groundwork as local SEO), and third-party presence you did not write.
The third is the slowest and hardest. We would rather say that than sell a package that quietly omits it.
Results
As of October 10, 2026, we have not published a completed, controlled citation study for Novarius. This page describes the method and does not claim a citation rate or client results. A future report should include its question set, settings, dates and evidence, including absences.
How to run this yourself
Keep one record per question, engine and run:
- Write 12–20 questions a customer might actually ask. Include brand, service, problem and local.
- Open a logged-out browser session.
- Ask each system each question.
- Record the answer, the sources cited, and whether you appear.
- Screenshot everything.
- Repeat monthly with the same questions and record any changes in settings.
Our AI search optimization assessment covers access, content, identity and observed citations. Scope and repeat monitoring are agreed before work starts. Repeated observations can show a trend; they do not prove which change caused it.