How SELISE GEO measures AI visibility

1. Your brand, read from your own sites

You add a brand by name, with its website if it has one and every other place it shows up: more websites, LinkedIn, Instagram, YouTube, Medium and so on. SELISE GEO reads what it can of these places and drafts a brand profile: what the brand mainly is, the lines of business it sells (named the way buyers search for them, never by product names), what sets it apart, and the markets it works in. Nothing is saved until you confirm it.

2. The people who buy from you

Visibility only matters for the questions real buyers ask. SELISE GEO drafts personas, specific customers of your brand, each with a short description, their needs, the market they live in and the language they type in. You name them, edit them and keep the ones that are right.

3. The questions they ask

From your offer lines and your personas' needs, SELISE GEO drafts topics and then questions, each written from one persona's point of view. Every question passes an eligibility check: would this person really ask it, could your brand be a fair answer, is it on topic. Questions that name your brand are tracked as branded, all others as organic. You accept the questions you want to track; they become the stable baseline every audit re-runs.

4. The audit

An audit has two parts:

• The engine fan-out. Every active question goes to every AI engine on your plan, with web search where the engine supports it, and the full answer is stored together with every page the engine cited or read.

• The site audit. SELISE GEO checks your website the way AI crawlers see it: whether each AI crawler may read your homepage and key pages under your robots.txt rules, and the technical signals that help or block them.

5. How each answer is graded

Every answer is read for:

• Citation strength: whether your brand appears and how prominently, scored 0, 25, 50, 75 or 100.

• Sentiment: whether the mention is positive, neutral or negative.

• Position: where your brand ranks in a list of recommendations, next to every competitor named.

• Sources: which websites the engine relied on, and whether your own site is among them.

A grade crediting your brand is overruled when the answer shows no evidence of it: its name, an alias, its domain or its website among the cited pages. The text decides whether your brand is there; the grading decides how prominently.

6. The visibility score

The visibility score combines all graded answers of a run:

• 70% comes from organic questions, the ones that do not name your brand, because being recommended unprompted is what wins new buyers.

• 30% comes from branded questions, which show whether the engines know your brand when asked about it directly.

• An answer that recommends against your brand scores 0.

The same score is shown per engine. Share of voice, positions and coverage are measured on organic questions.

7. From score to action

Each completed audit produces:

• Gaps: the organic questions where no engine cited you, and who was recommended instead.

• Sources: the websites behind the answers, by type (your own, competitors', editorial, review, directory, community, reference), and where your brand is missing.

• Actions: a ranked worklist of site fixes, AI-crawler access, content to publish for topics where you are absent, and third-party sites to be mentioned or reviewed on.

8. Tracking over time

Re-running the same baseline shows how your visibility moves, engine by engine, and which actions confirmed as fixed. When you change your personas or questions, later runs are flagged so you never compare unlike with unlike.

Limits worth knowing

AI answers vary between engines, models and points in time. Scores are a consistent measurement for comparison, not a permanent truth, and no tool can guarantee that an engine will cite a brand. See the Terms of Use for details.