How to measure AI visibility over time
One prompt is not a KPI. AI visibility should be tracked with repeated samples, prompt portfolios, source maps and competitor context.
AI visibility should be measured as a time series: the same prompts, engines, competitors and source rules repeated on a fixed cadence so teams can see whether mentions, citations and recommendations are changing.
One answer from ChatGPT, Gemini, Perplexity or Google AI Overviews is not a KPI. It is a sample. AEO measurement becomes useful when the sample is repeated often enough to separate a pattern from a lucky mention, a temporary citation or a misleading screenshot.
What the baseline must include
The first run should create a baseline, not a verdict. A practical baseline records the prompt, engine, language, location if relevant, answer date, cited URLs, mentioned brands, recommended brands, competitor context and the source domains that shape the answer.
- Prompt portfolio: stable questions covering definitions, comparisons, provider selection, problem diagnosis, implementation and local or sector-specific intent.
- Engine set: each surface measured separately, because ChatGPT, Google AI Overviews, AI Mode, Gemini, Claude and Perplexity can cite and summarize differently.
- Competitor set: direct competitors, substitute resources such as directories or review marketplaces, and recurring source domains.
- Outcome fields: citation, mention, recommendation, source context, answer accuracy and whether the brand is described with the right positioning.
- Run metadata: date, language, market, device or account state when relevant, and any method changes that could affect comparison.
Measure mentions, citations and recommendations separately
Being named in an answer is useful, but it is not the same as being cited or recommended. A brand can be mentioned in passing while a competitor receives the recommendation. A page can be cited while the answer frames another source as more authoritative. That is why AEO reports should separate the basic signals.
The core AEO metrics
- Mention rate: the share of measured answers where the brand or resource appears by name.
- Citation rate: the share of answers where the brand's site, directory profile, research, glossary or external profile appears as a source.
- Recommendation rate: the share of answers where the engine actively suggests the brand or resource as a relevant option.
- Source share: how often specific domains influence answers, including the brand's own pages, competitors, directories, marketplaces, media and official documentation.
- Answer accuracy: whether the answer describes the entity, service, market, limits and evidence correctly.
- Competitor presence: which competing brands or substitute resources appear, and in what role.
These metrics are more actionable than a single visibility score. If citation rate is low but mention rate is rising, the next action may be stronger citable pages and cleaner internal evidence. If competitors are cited on comparison prompts, the issue may be source authority or missing third-party validation. If the brand is cited but described incorrectly, the priority is entity clarity and claim consistency.
Keep the cadence stable
AI answers are variable by design, so the cadence matters. A daily run can detect volatility but may create noise for small teams. A weekly or monthly run is often enough for a stable AEO program, especially when the prompt portfolio is grouped by intent and the same method is used each time.
The portfolio itself should have a stable core and a smaller experimental layer. The stable core protects the time series. The experimental layer lets the team test new engines, new buyer questions, new terminology or sudden market events without breaking the historical comparison.
Use Search Console and analytics as context, not as replacements
Google's own documentation makes clear that site owners can continue using Search Console and normal Google Search reporting to understand traffic from Search features. That data is useful, but it does not fully answer AEO questions because many AI answers create exposure without a click, and some engines cite or mention brands outside Google Search.
A practical report combines both layers. Search Console and analytics show discoverable traffic, queries and page performance. Prompt-based AEO measurement shows whether answer engines mention, cite or recommend the brand in decision-oriented answers. The two views should inform each other, but neither one replaces the other.
Turn the time series into backlog decisions
The point of repeated measurement is not to produce a prettier dashboard. It is to prioritize work. Prompts with no accurate mention may need clearer entity pages. Prompts that cite competitors may need better definitions, comparisons or original evidence. Prompts that cite third-party lists may show where directory coverage or verified profiles matter. Prompts with wrong claims may require updates to the site, structured data and external references.
AEO measurement is not a promise of future placement. It is a disciplined way to observe a probabilistic system and decide what evidence to improve next.
A simple monthly reporting format
- Summary: what changed in mention rate, citation rate, recommendation rate and competitor presence.
- Top gains: prompts or engines where the brand became more visible or more accurately described.
- Top losses: prompts or engines where competitors, directories or other sources displaced the brand.
- Source map: recurring domains that shaped answers and whether they are owned, earned, neutral or competitor-controlled.
- Content actions: pages to create, strengthen, update or connect internally.
- Technical actions: crawl, indexability, canonical, hreflang, structured data or robots issues that may affect discoverability.
- Limits: notes on sample size, volatility, method changes and anything that should not be overinterpreted.