What we do, concretely: the seven families of interventions

A GEO plan is not a list of generic tips: it is the list of interventions that change the sources from which AI engines build their answers about you. The interventions fall into seven families. Not all of them apply to everyone: the action plan of each assessment says which ones matter in your case, in what order and with what expected effect.

1. Entities and primary sources

Models must first of all understand who you are: exact name, industry, headquarters, size, key people, and not confuse you with namesakes. The sources they use for this are few and well known:

• Wikidata: the open knowledge base that feeds Google's Knowledge Graph and is read by all models. Creating or completing the company's entry (identifiers, headquarters, industry, website, profiles) is almost always the first intervention.
• Wikipedia: only if the company meets the notability criteria. An existing entry frozen for years is one of the most frequent causes of dated narratives.
• LinkedIn: complete company page consistent with the website, aligned leadership profiles.
• Local listings and registries: Google Business Profile, Bing Places, company registers, industry registries, category directories (for B2B services, agency and supplier directories matter a great deal).

Expected effect: answers that identify you correctly and a stable entity to which models can attach the subsequent facts.

Included in the assessment: Wikidata entry created or completed, LinkedIn page aligned, Google Business and Bing Places checked, consistent data across all sources. With the GEO Program: a new Wikipedia article, registers and directories beyond the first three.

2. Extractable proprietary content

AI engines do not read pages the way people do: they extract self-contained passages. A website full of advertising claims and poor in facts offers nothing to extract. We intervene with:

• a "facts and figures" page (founding, size, sites, certifications, clients, awards) updated with a visible date;
• definitions and FAQs at the top of key pages, written to be cited: question, short answer, detail;
• data with source and date, quotations from real people, tables: studies show that statistics and quotations significantly increase the probability of being picked up;
• placing facts in the first third of the page, where models take most of their passages.

Included in the assessment: a "facts and figures" page and the FAQ of the key page, drafted, reviewed by us and ready to publish. With the GEO Program: articles, white papers, case studies, microsites and topic hubs, and the revision of existing pages.

3. Structured data

Schema.org data (Organization, Person, Product, Service, FAQPage, Article) tell machines what is on the page and link it to the entity: same names, same identifiers, same profiles (sameAs) everywhere. We check consistency of name, address and contacts across all sources ("NAP consistency") and add the missing nodes.

Included in the assessment: Organization and Person JSON-LD with sameAs, FAQPage schema for the FAQ delivered, ready to upload. With the GEO Program: Product, Service and Article across the whole catalogue or blog.

4. Third-party sources

This is the family with the greatest weight, because models trust what others say about you more than what you say yourself. Depending on the audience:

• press and trade publications: interviews, contributed articles, releases with original data (an observatory, a study) that outlets pick up and AI cites;
• reviews: Glassdoor and Indeed for candidates, category platforms for products and services. It is not about "managing" reviews but about replying, updating and inviting those who are satisfied;
• communities: Reddit and industry forums are among the sources ChatGPT and Perplexity cite most. Take part with useful content, never promotional;
• directories and rankings: for many B2B categories AI cites the same three or four directories. Being there, with a complete listing, changes category answers.

Included in the assessment: a complete listing on up to three category directories indicated by the plan. With the GEO Program: press and interviews, review campaigns, community and forum presence, ongoing LinkedIn posts.

5. Correcting dated narratives

The most common cause of wrong answers is not a false source: it is a true but old source, contradicted by nothing more recent. A closed lawsuit, a restructuring years ago, a discontinued product. It is corrected with:

• updating the primary sources (Wikipedia, Wikidata, company listing) with the subsequent facts;
• a "what has changed" page on the website, with dates and sources, that models can read;
• releases and articles reporting the current state, so that a more recent, more authoritative source exists;
• public replies where the narrative was born (reviews, communities).

Included in the assessment: Wikidata and listings updated, one documented proposal on the Wikipedia talk page, a "what has changed" page drafted if the source map calls for it. With the GEO Program: press releases, articles and public replies over time.

6. Technical aspects

The previous interventions are useless if machines cannot read the pages:

• AI crawler access: many hosts and CDNs block GPTBot, ClaudeBot and PerplexityBot by default. We check robots.txt, firewalls and anti-bot rules;
• Bing: ChatGPT and Copilot read the Bing index. Property on Bing Webmaster Tools, sitemap submitted, IndexNow active;
• Google Search Console, sitemap, canonicals, speed, clean HTML with no content hidden behind scripts.

Included in the assessment: AI crawlability audit of the site (16 crawlers, 10 checks), robots.txt, llms.txt, sitemap and IndexNow ready to upload, Bing Webmaster Tools and Search Console properties with the sitemap submitted. With access to the site we apply them ourselves. With the GEO Program: work on the CMS, speed and site architecture.

7. Monitoring

No intervention is final: models update, sources change, competitors move. The 90-day re-analysis is included in every assessment; with the GEO Program the run is monthly, with alerts on negative narratives and a quarterly before/after report.

Who does what

The baseline actions are included in the assessment price and we carry them out: the one-off actions, with predictable effort, on the company's own channels and on open primary sources. Everything that requires serial content production, relationships with third parties or ongoing presence is the GEO Program, quoted on the plan. In between sits your team, which can carry out any action in the plan with our instructions.

Details and price are on the Pricing page.