Help Understanding your results
How you appear in AI answers
Read the heart of the report: which questions were asked, whether you were named, and in what tone.
AI visibility slides overview
This block is what the whole audit exists for. It opens with two short explainer slides — how the scan works, and why the language of the question matters — then moves to your own results.
Three slides carry your data: your visibility split by kind of question, then every single prompt with its result, then the tone of the answers where you appeared.
The per-prompt slide is the one nobody should skip. Percentages are summaries; the prompts are the actual questions your audit asked, and reading them is the fastest way to know whether you’re measuring the right thing.
Why the language slide is there
Assistants answer in the language they were asked in, and they answer from sources in that language. Ask about a Japanese market in English and you get a different answer, from different sources, about different companies.
The slide explains that your prompts are asked in your market’s language — the one your buyers use — which may not be the language of your website or of this report.
AI visibility by funnel stage
Your visibility split two ways.
By funnel stage — three kinds of question, and they mean very different things:
| Stage | The question behind it |
|---|---|
| Awareness | Someone with a problem who doesn’t know you, or anyone else, yet |
| Comparison & choice | Someone weighing options against each other |
| Decision · by brand name | Someone already asking about you, by name |
Prompts you wrote yourself that don’t fit a stage appear as No stage (manual prompts).
Each stage shows how many prompts sit in it and what share of your total that is. A stage carrying two prompts is not a measurement of that stage — it’s an anecdote.
Branded versus unbranded — the second cut, and the one that changes decisions. Branded queries already name you; unbranded ones are where buyers choose among suppliers they don’t know.
Good to know Strong on branded, weak on unbranded is the most common pattern there is, and it is not good news wearing a hat. It means AI confirms you to people who already know you, and leaves you off the list for everyone else. Improving it is content work, not technical work.
AI-answer visibility by prompt
Every tracked prompt, with six columns:
| Column | What it tells you |
|---|---|
| Visibility | How often you were named in the answers to this prompt |
| Tone | Whether what was said about you was positive, neutral or negative |
| Competitors | Who else appeared in those same answers |
| Position | Where you sat in the answer — lower is better |
| Sources | How many distinct domains the assistants cited |
| Topic | Which of your topics the prompt belongs to |
Long lists run over several slides; each says “Part N of N.”.
Read the prompt text itself first, before the numbers. If a prompt isn’t how your customers would actually ask the question, its score is measuring the wrong thing — and prompts are yours to change, in the app. See Prompts and topics.
A prompt where competitors appear and you don’t is the most actionable row in the entire report.
Sentiment in AI answers about you
Four parts, all of them counting only the answers where you were actually mentioned. That’s stated on the slide, and it’s what makes the numbers meaningful: tone across answers that never named you would be noise.
- Sentiment distribution — the split of positive, neutral and negative, with the average, and how many prompts with mentions it’s based on.
- By engine — average tone in each assistant. Big gaps between engines usually mean one engine is reading a source the others aren’t.
- Strong prompts — where you’re spoken of best.
- Weak prompts — where the tone is lowest. Start here.
When nothing negative was found, the slide says so rather than drawing an empty chart.
Tone isn’t something you edit. It changes when what’s published about you changes — see Sentiment.