What we do, what is included, what it costs: the GEO Assessment on one page

The GEO Assessment is one service. We measure what AI engines answer about a company, verify where the answer comes from, write the plan, carry out the baseline actions ourselves and measure again after 90 days. The price is public, depends on the scope and sits at the bottom of this page. The first step is a free audit, in 48 hours.

Why it matters

Whoever evaluates a company, a candidate, a buyer, an analyst, a journalist, increasingly asks an AI before searching Google or writing. The AI answers with what it finds in the sources it can read. If a company does not appear, or appears with an outdated narrative, other sources compose the answer.

The data sources, and what happens to those who are not there, are in Why now.

What we do

1. We measure. We build the questions the audience you care about really asks AI engines, category questions and questions about the company's name, and put them to five engines (ChatGPT, Gemini, Perplexity, Claude, Copilot) with the ReputAction platform.
2. We verify the sources. For every answer we trace the pages that generated it: website, Wikipedia and Wikidata, press, reviews, LinkedIn, communities. URL by URL, by hand.
3. We write the plan and do the basics. Actions by channel, in order of priority, with who does what. The baseline actions we carry out ourselves, included in the price; the rest is done by your team with our instructions, or by us with the GEO Program.
4. We measure again. After 90 days we repeat the same set of questions on the same engines. The before-and-after comparison is part of the work, not an option.

The method in detail is in How it works; the families of intervention in What we do.

Why ten questions to ChatGPT are not enough

Anyone can open ChatGPT and ask "what do you know about this company?". It is a good first test, and we suggest it in the DIY test. But what you get is an opinion: one answer, from one engine, to one wording, at one moment. An assessment is a measurement. The difference lies in five points.

• AI engines never answer the same way twice. They are probabilistic systems: the same question, reworded or repeated, changes the answer. Ten questions by hand cannot tell a one-off from a pattern. The system puts the same set in different wordings, from the point of view of different profiles, to five engines, and measures frequency: how often the company appears, in what position, with what tone.
• The questions that matter do not name the company. The point is not what the AI says when you ask about you, but whether it cites you when a candidate asks "where is it worth working in this industry" or a buyer asks "which suppliers would you recommend for…". These are category questions, by topic and by profile, and they have to be built in series: they do not occur to someone testing by hand, and by hand they cannot be repeated identically.
• Seven families of probes, for every engine and every topic. Overall picture, strengths, weaknesses, what the engine really knows and from where, spontaneous visibility against competitors, real questions simulated by profile, intervention priorities. Up to twelve topics, five engines, answers in a structured, comparable format. In one run the system formulates and evaluates hundreds of questions: it is a sample, not a chat.
• Verified sources, not declared ones. AI engines cite sources that sometimes do not exist or do not say what is attributed to them. The system retrieves the real sources with a search engine, compares them with the declared ones and keeps the URLs: the map comes from there, and the plan from the map.
• Repeatability. The same set, after 90 days, on the same engines: only then is the before-and-after comparable and the result of an intervention demonstrable. Ten questions asked by hand cannot be repeated identically even the next day.

What you get

• GEO Score per engine and overall, from A to F, comparable over time.
• The real answers, verbatim, question by question and engine by engine.
• The source map with the verified URLs that weigh on the answer, and those that are missing.
• Risks and outdated narratives: what the AI engines say that is wrong, old or negative, and where they take it from.
• The competitor comparison: the ones you name and the ones the AI engines cite instead of you.
• The action plan by channel, with priorities, and the debrief call with your team.
• The baseline actions carried out, listed below.
• The 90-day re-analysis, with the same set of questions.

The report lives in the platform and can be exported. The indicators we use and two anonymised cases are in Results.

What is included in the price

A report that only recommends leaves the work to whoever receives it. That is why the price includes the baseline actions: the one-off actions, with predictable effort, on the company's own channels and on open primary sources. We carry them out within six weeks of the debrief call, so the 90-day re-analysis measures their effect.

Entities and primary sources

• Wikidata entry created or completed: identifiers, headquarters, industry, website, profiles, key people;
• LinkedIn company page aligned in its texts: description, industry, location, website;
• Google Business Profile and Bing Places checked or updated;
• consistent name, address and contacts across all sources.

Technical

• AI crawlability audit of the site: 16 crawlers, 10 checks, score from A to F;
• robots.txt, llms.txt, sitemap and IndexNow ready to upload; with access we apply them ourselves;
• Bing Webmaster Tools and Google Search Console properties, with the sitemap submitted.

Extractable content, drafted, reviewed by us and ready for the site

• "facts and figures" page: founding, size, locations, certifications, clients, awards, with a date;
• FAQ for the key page of the chosen audience, with FAQPage schema;
• Organization and Person JSON-LD with sameAs;
• "what has changed" page, if the source map reveals outdated narratives.

Third-party sources, one-off

• one documented proposal on Wikipedia, on the talk page and with sources;
• a complete listing on up to three category directories indicated by the plan.

90-day re-analysis, with the same set of questions and the before-and-after comparison.

What is not included

Everything that requires serial content production, relationships with third parties or ongoing presence, and therefore has no predictable effort before reading the plan: microsites and topic hubs; articles, white papers and case studies in series; relationships with the press, interviews, distributed press releases; review campaigns on Glassdoor, Indeed or category platforms; Reddit and forum presence; ongoing LinkedIn posts; a new Wikipedia article; life sciences content requiring medical-regulatory review; additional audiences, countries or languages; monthly monitoring with alerts. That is the scope of the GEO Program, quoted on the concrete plan, or of your team with our instructions.

What it costs

A single starting price, public, so that whoever writes to us knows the order of magnitude. The scope is defined together, after the free audit.

The exact scope (audiences, countries, competitors, number of questions) is agreed on a call: the free audit is also for that.