Our measurement method
Evidence you can inspect.
A visibility score is useful when you know what was measured. Akkrev keeps the questions, answers, model names and cited sources together, so you can check the result.
Questions a guest would ask
Your venue type, location and owner-supplied facts guide a set of 25, 50 or 100 guest questions. They cover needs such as family stays, dietary options, neighbourhoods and occasions. You can add or remove questions in your dashboard. Questions are samples of possible guest needs, not search-volume estimates.
Four model families, once a week
We sample models in the OpenAI, Google Gemini, xAI Grok and Anthropic Claude families. Included Grok answers come from model memory without live web search. The other three families use web search. Grok therefore does not provide fresh web evidence for its recommendations. These sampled API answers can differ from ChatGPT, Gemini, Grok and Claude's consumer apps. They do not measure Google AI Overviews or every personalised answer a guest might see.
The actual model ID and returned source URLs are saved with each answer. The source links returned with each answer show the evidence available for that sample. A cited source is evidence of what that sample used, not a complete list of everything that influenced a model.
Optional model coverage
Paid Grok search uses web search for the weekly questions covered by your paid tier. Extra model add-ons use the existing weekly questions without web search. Changes apply to new weekly reports and start a new comparison baseline. Original answers and CSV exports preserve each sample’s method. A base-plan upgrade does not automatically increase a lower-tier Grok search allowance.
How the score works
Your score is the percentage of usable sampled answers that recommend your venue. If it appears in 30 of 100 usable answers, the score is 30%. It is not a market share, a Google ranking, or the percentage of real guests who will find you.
Temporary failures retry automatically. Failed answers are excluded from the denominator and partial coverage is labelled. A complete collection failure displays no score. You can inspect full answers and download a CSV.
Weekly changes are shown only when the question set, sampling configuration and actual model mix match and both reports have full coverage. Changes to questions, models or web search settings start a new comparison baseline. Even comparable samples fluctuate. One weekly change does not establish that a particular edit caused it.
From gaps to useful content
Reports suggest three next steps based on missed questions and cited sources. On Grow, two original article drafts are prepared each month. Scale includes four. Drafts use the facts you supply. Review them for accuracy, then publish to a connected WordPress website, or copy them into another website editor. Shopify publishing and Lovable change preparation are not available yet. See supported website connections for capabilities and account requirements.
Automatic publishing is off by default. Turn it on explicitly after connecting your website if you want future articles to publish without individual approval. Your website connection is encrypted. You remain responsible for your venue facts and content. Akkrev does not change your Google Business Profile, reviews or third-party listings automatically.
Good SEO still matters
Make pages useful, crawlable and easy to navigate. Keep venue details consistent, answer real guest questions and use structured data that matches what visitors can see. No special AI markup or llms.txt file guarantees inclusion. The optional venue facts download is a convenience, not a ranking requirement.
Google's guidance for AI search features. Google's scaled content policy
Google's observations and our samples
Search Console's Generative AI features report shows impressions from Google's AI Overviews and AI Mode. Those impressions are also included in overall Google Search performance. They are a different measurement from our sampled API recommendation rate. The Google search results view in Akkrev currently reads overall Web search data; it does not import the dedicated generative AI report or identify which visit produced a booking.
Use Google's generative AI report alongside our practical measurement guide. Do not add its impressions to overall Search impressions as though they represented a separate audience.
Measure business outcomes separately
Use your website analytics and booking system to follow referral visits, enquiries and bookings. A mention in a model answer does not prove a visit or a booking. Referral data can be incomplete, so ask guests how they found you as an additional signal. We do not promise rankings, recommendations, traffic or revenue.