How to read your report
Understand the evidence behind an AI visibility report.
KnewSearch's free scan gives you a modeled starting assessment of one brand's visibility in AI answers. Use it to identify questions worth checking and marketing work worth investigating.
How the free baseline is made
The scan starts with your domain and any competitors you provide. It attempts to read your homepage title, description and a short text excerpt to understand the business. This is a limited view of your website, not a crawl of every page.
An AI model uses that context to suggest buyer questions, assess the brand and named competitors, and propose priorities. The scan can use search-grounded analysis when available. It can also fall back to model knowledge, so a completed report does not guarantee that a fresh web search informed every finding.
Google describes search grounding as connecting an API response to web information. That can improve the context for an assessment, but it does not turn one model's assessment into separate observed answers from four consumer products. Read Google's explanation of search grounding.
Read the score as an assessment
The free scan's score, tiers, competitor rankings and recommendations are model-generated estimates. A score of 60 out of 100 does not mean your domain appeared in 60% of buyer conversations. Suggested questions are ideas to investigate, not a verified search-volume dataset or a completed tracked-question count.
Engine names in a modeled report identify the products being assessed. They do not establish that KnewSearch opened each product and observed the answer a buyer would see. Treat statements about likely gaps and competitors as hypotheses until supported by a relevant observation.
What you can conclude
Identify the kind of result before deciding what to do with it.
- Modeled assessment
Choose buyer questions and possible gaps to investigate.
An estimate does not establish an observed brand mention or citation.
- Domain found in observed sources
Record a source association for that question, collection method, place and time.
A source link does not prove that the answer recommended the brand or sent it traffic.
- Completed check, no matching source domain
Report that the domain was not found in this observation's returned sources.
This does not establish absence from the answer text or from other questions.
- Unavailable observation
Mark the result unknown and record which evidence is missing.
A failed or missing response cannot be counted as a completed check with no matching domain.
What a citation observation means
A citation observation records a domain found in source references returned for a particular question and provider response. It is narrower than a claim that the provider recommends the brand, agrees with its claims, ranks it first or sends it traffic.
The September 10 sample uses DataForSEO's Google AI Overview collection, ChatGPT web scraper and Gemini web scraper for one question in United States English. Its source lists contain references associated with the returned answer blocks. They do not establish exact inline citation spans. These are web observations collected through a third party, with the method and time shown alongside each result.
Other collection methods
Direct provider APIs are separate interfaces from the sample's web collection methods. OpenAI documents citation annotations associated with an answer. Perplexity distinguishes referenced citation URLs from the wider search results supplied as context. Google's grounding metadata identifies web sources and links parts of an answer to them. These differences matter when interpreting a report.
A provider API response is a separate observation from a session in the consumer app. The model, instructions, search settings and conversation can differ. Record the method used and keep those observations separate rather than presenting them as interchangeable.
Missing evidence has a meaning of its own
An answer with no returned citation domains establishes only what that response returned. A failed request or unavailable provider supplies no answer to judge. Neither supports a claim that your brand never appears in AI search. DataForSEO also documents that a web-enabled response can contain no citation annotations. Read DataForSEO's response documentation.
Results can vary. A different question, provider or comparison group changes what is being evaluated. A changed modeled score is not proof that a marketing action caused an improvement.
Turn the baseline into a useful next step
Start with one domain and one marketing decision. Review the suggested questions, choose a relevant page or campaign, and identify the evidence you would need before acting. Separate the assessment, any observed result and your recommendation when sharing the report.
This approach fits in-house marketers, SEO and content teams, consultants and agencies across industries. It is a starting point for investigation, not a promise of citations or rankings.
To inspect a dated observation from our own website, open the sample report and its source pages.