Methodology statement

This statement describes how we build an AI Visibility Report: what we test, how we collect data, how we score results, and the limits of what the numbers mean. It forms part of our Terms of Service.

1. What we measure

An AI Visibility Report answers one question: how visible is your business when people ask AI search engines for a recommendation in your category and location?

We test that visibility across six AI engines: ChatGPT (OpenAI), Google AI Overview, Perplexity, Gemini (Google), Claude (Anthropic), and Grok (xAI). For each engine and each query, we record whether your business was named, recommended, or cited, and which other businesses appeared instead.

Two service levels are available:

2. How a report is produced

Every report follows the same sequence of steps, run automatically at the time you purchase.

Query generation

We build a set of natural-language questions that a real customer would ask, based on your business category and location. The free scan runs 3 queries; the full report runs 10. The queries are designed to reflect how people actually search, not just how businesses describe themselves.

AI engine testing

Each query is submitted to each AI engine through official commercial access. We record the engine’s response — specifically whether your business was named, recommended, or cited, and what other businesses and sources appeared. Nothing is read into or inferred beyond what the engine returned.

Website analysis

We read your business’s own public website to understand how well it gives AI engines the clear, structured, answer-shaped content they look for when deciding what to recommend.

Public business signals

We collect business-level facts from public sources: your publicly listed rating and review aggregate, directory and social profile presence, and other signals that affect how much verifiable evidence AI engines have to draw on. See section 4 for a full breakdown of source types.

Report generation

All findings are combined and the report is written by an automated AI workflow. Reports are AI-generated and are not reviewed by a human before delivery. This is disclosed at checkout and in every report.

3. How we score

Results are combined into a score out of 100 across four pillars. Each pillar captures a distinct dimension of AI visibility.

AI Presence

How often, and how prominently, the AI engines named, recommended, or cited your business across the tested queries. This is the direct output of the engine tests: it is not modelled or estimated.

Content Quality and Depth

How well your own website and content give the engines clear, extractable, answer-shaped material to cite. AI engines prefer content that directly answers questions; this pillar assesses how far your existing content does that.

Authority and Trust

The strength of third-party evidence that makes your business easier for an engine to recommend with confidence: reviews, directory listings, mentions, and other independent signals that corroborate who you are and what you do.

Technical Readiness

Whether your website’s technical foundations allow AI engines to read, understand, and reference it correctly: crawlability, page speed, structured data (schema markup), and related technical factors.

A score measures the evidence available to AI engines at the time of the scan, not the quality of your business. A lower score means the engines had limited verifiable evidence to draw on, not that your business is poor. Improving your score means giving the engines more of what they need to recommend you.

4. Data sources

We prefer official APIs and licensed data feeds wherever they exist. We do not scrape platforms that prohibit automated access. The table below lists each type of source, what we use it for, and what we collect.

A full list of our data sub-processors is published in our Privacy Policy.

5. What we do not collect

Our reports describe businesses, not people. Before any third-party data is stored or included in a report, we apply a strict filter. We never retain or display:

• Individual names — staff, practitioners, owners, or reviewers found on competitor pages.
• Individual review text — we collect aggregate ratings and counts only, never the content of specific reviews.
• Personal contact details — no email addresses, phone numbers, or home addresses.
• Profile photos or personal social handles.

Where a business trades under a person’s name, we use that name only as the business entity label and attach no personal information about the individual.

For full details on how personal data is handled, see our Privacy Policy.

6. How competitor data is presented

Your report includes data about competitor businesses that the AI engines surfaced in response to your queries. We present this data as neutral, sourced, point-in-time observation, not a ranking or quality judgement.

Specifically:

• We report what the engines said — which businesses were cited and how often, based on the specific queries we ran.
• We describe differences in publicly available evidence patterns (for example, review volume, structured data presence, broader content coverage) rather than making quality judgements about a named business.
• We do not publish competitor star ratings or individual review text.
• Competitor data reflects a sampled diagnostic of the specific queries run at the time of your scan, not a comprehensive industry survey.

Any business that appears in a report may request a correction or removal by contacting hello@getrecommended.io.

7. Regulated health businesses

Where a business is a regulated health service, our recommendations are constrained. We do not advise soliciting or displaying patient testimonials, patient reviews, outcome claims, case studies, or before-and-after results. These are prohibited in advertising for regulated health services under the Australian National Law (administered by AHPRA) and equivalent rules in other markets.

For regulated health businesses, acceptable authority signals are limited to credentials, registration, professional-association membership, published methods, peer-reviewed research references, and organisational authority signals.

8. Limitations of this method

We are transparent about what this report can and cannot tell you.

• Point-in-time only. AI engines update their models, training data, and responses constantly. A report reflects the state of AI search at the moment of your scan. Results may differ if the report is re-run.
• Sampled queries, not exhaustive. We test a defined set of queries, not every possible phrasing a customer might use. A business may perform differently on untested queries.
• Model non-determinism. The same query can produce different answers on different runs. We record what was returned at scan time.
• Google AI surfaces. Google AI Overview and AI Mode are accessed through a licensed search-data provider; Google does not offer a public API for these surfaces. Coverage depends on that provider.
• Third-party data may be stale. Review aggregates and profile data reflect what the source published at scan time. We record a source and timestamp for every data point and provide a correction route for any factual error.
• Scores measure evidence, not quality. A mention count, rating aggregate, or pillar score is a signal of visible evidence available to AI engines, not a verdict on how good a business is.

9. Corrections and questions

To request a correction or removal of any business data that appears in a report, or to ask how a specific number was produced, contact us at hello@getrecommended.io.

For how we handle personal data, see our Privacy Policy. For the terms governing the report, see our Terms of Service.