How the GEO Assessment works: measure, understand the sources, act, measure again

The GEO Assessment is a repeatable analysis of what generative AI engines say about a company, run with a fixed set of questions on several engines, that traces the sources of every answer and turns them into a measurable action plan. Repeating the same analysis after the interventions shows what has moved.

The method has four phases. The first two produce the picture; the third changes things; the fourth proves it.

Phase 1. Measure: real questions, on the right engines

We start from the audience, not the website. For each audience (candidates, customers, buyers, stakeholders, investors) we build a set of questions that reproduces what those people actually ask the AI:

• branded questions: "What is it like to work at [Company]?", "Is [Company] a reliable supplier?", "Who is [CEO]?";
• category questions, where your name should appear but might not: "Which are the best [industry] companies in [country]?", "Which [product] should I choose for [need]?";
• comparison questions: "[Company] or [Competitor]: where should I apply / buy / invest?";
• risk questions: "What controversies has [Company] had?", "Are there problems with [product]?".

The questions are put, in the agreed language and country, to five engines: ChatGPT, Gemini (with AI Overviews), Perplexity, Claude and Copilot. Each engine answers differently because it reads different sources: that is why an analysis on a single engine is partial.

For each answer we record: whether the company is present, in what position, with what sentiment, which facts are reported and whether they are correct and current, which competitors appear next to you.

Phase 2. Understand the sources: where every sentence comes from

This is the part that distinguishes an assessment from a simple test. For every relevant statement in the answers we trace the source that generated it: the article, the Wikipedia page, the review, the discussion, the listing. Each source is verified with URL, date and estimated weight.

The result is the source map: the list of pages that are shaping your perception in AI, ordered by influence, with an indication of what is dated, incomplete or missing. In most cases the map shows two things: that the decisive sources are not the ones the company looks after, and that current strengths appear nowhere in a machine-readable form.

Phase 3. Act: the plan, and who carries it out

From the comparison between answers and sources comes the action plan: a list of interventions by channel, with priority and expected impact. There are seven families of intervention, described on the What we do page: entities and primary sources, extractable proprietary content, structured data, third-party sources, correction of dated narratives, technical aspects, monitoring.

The plan's baseline actions we carry out ourselves, included in the price: entities and listings, crawlability and technical files, structured data, FAQ and facts page drafted, one Wikipedia proposal, the first directories. The rest is carried out by your team with our instructions, or by us with the GEO Program. Either way every action has an owner, a date and a way to measure its effect. The line between included and not included is on the Pricing page.

Phase 4. Measure again: the same analysis, 90 days later

The assessment includes a 90-day re-analysis with exactly the same set of questions, on the same engines. It is the only honest way to measure: changing the questions would change the results. The before/after comparison shows, engine by engine, what has moved: presence, accuracy, sentiment, competitors cited next to you. The baseline actions are completed within six weeks of the call, so the re-analysis measures their effect.

With continuous monitoring the run becomes monthly and alerts are added: if a negative narrative emerges, you know before it becomes the standard answer.

What the report contains

The report lives in the ReputAction platform and is exportable to a branded PDF, ready for the board. It contains:

• GEO Score from A to F (0-100), a summary of presence and answer quality, comparable over time and with competitors;
• KPIs: presence in answers, average sentiment, number of critical sources not covered, median year of cited sources;
• visual SWOT: strengths, weaknesses, opportunities and risks as they emerge from the answers;
• sample answers with highlighted passages and the source of each;
• source map with verified URLs and weight;
• reputational risks ordered by severity and likelihood;
• action plan by channel, with priority and owner;
• competitor comparison on the same questions.

An example of the report, with fictional data, is visible in the "Inside the platform" section of every assessment.

What we need from you

Little: the name of the company and brands, countries and languages, the audiences that matter, two or three reference competitors. No access to your systems is needed. For the free audit the contact form is enough.