Features
AI visibility and engines
The AI visibility view brings together your tracked prompts, the AI engines' answers and your site's “AI-ready” checks. This page describes every card and action; the measurement method is in the AEO page.
What you need to measure
- A plan that includes AI: observations use the organisation's monthly AI quota. The free plan includes none.
- At least one verified engine on the platform (see the header's “AI engines” pill).
- Tracked prompts: the default set created by an analysis, and your own.
Track a prompt and run an observation
- In AI visibility, enter the prompt (3 to 1,500 characters).
- Choose “All verified engines” or a single engine, then “Track this prompt”. A site can track 25 prompts at most from this card.
- Click “Run the observation”. The maximum cost is reserved on your quota before the call; if the monthly quota is used up, the observation is refused with a message saying so.
- The KPIs update when the run ends (finished, partial or failed, visible in Observations and alerts).
The trash icon removes a prompt from tracking. Answers already stored stay in the run history.
The KPIs
| KPI | Definition |
|---|---|
| Share of voice | Your brand's mentions ÷ mentions of your brand and tracked competitors, over the last three runs. See share of voice. |
| Brand mentions | Answers naming your site (name or domain, whole words). |
| Answers observed | Answers obtained from the engines over the period read. |
| Pages cited / prompts cited | URLs on your domain written in answers, and prompts with at least one answer citing you. |
| Verified engines | Engines actually queried, out of the engines known. |
| AI-ready | Summary of the robots, /llms.txt and Schema.org tiles below. |
The engine tiles
One tile per engine: ChatGPT, Gemini, Claude, Perplexity, DeepSeek, plus Grok (“activating” until it is enabled) and Meta AI where it is enabled, and Google's AI Overview, which NeoRank never queries itself. A tile's value is the rate of that engine's answers naming your brand — not a share of voice. Details opens the engine's panel: rate, pages cited, where to act and the list of uncited prompts.
The “AI-ready” tiles
| Tile | What is read | “Not measured” when |
|---|---|---|
| AI crawlers | Your live robots.txt, for GPTBot, PerplexityBot, ClaudeBot and Google-Extended | The file could not be read (4 s timeout, 15-min cache) |
| /llms.txt | Present: 200 answer as text or Markdown; absent: 404 or 410 | Any other answer, or site unreachable |
| Schema.org | Pages of the latest crawl carrying JSON-LD, and invalid blocks | No finished crawl |
Each tile opens a panel with an example and fix steps. The full robots reading (eight AI crawlers), the simulator and the generator are in the site audit; the llms.txt draft is generated from Settings › Accelerated indexing.
A prompt's resolution panel
On an uncited prompt, Resolve shows: the answers read per engine (excerpts), who cites or names you, the sites cited instead of you, what to aim for and the steps. It is the most concrete starting point: the page an engine cites instead of yours tells you what content it expected.
The sentiment card
The share of positive excerpts per engine and up to four quoted excerpts, classified by a French / English lexicon. With no classified excerpt: “Not measured”. Details in GEO.
Related views
- Query sets: each set, its share of answers mentioning you and the engine where you are most present. A set with no answer shows “Not measured”, never 0 %. Prompts are created from AI visibility.
- Digital PR and citations: the third-party domains cited, especially those cited without your site, and the share of each engine's answers citing them (not a trust score).
- Competitor analysis: AI citations per identified competitor (counted by domain) and mentions for tracked names.
- Observations and alerts: recent runs and their state.
Run states
| State | Meaning |
|---|---|
| Never measured | No observation has succeeded for this site yet. |
| Monthly quota used up | The month's AI budget is spent; it renews the following month (UTC). |
| Budget exceeded | The run's estimate exceeds the query set's budget. |
| Engine failed | One or more engines did not answer; the reason is shown. |
| Measured | The KPIs rest on real answers. |