GEO in practice: 7 concrete actions to improve AI visibility

Configure robots.txt to let AI bots read you

It's the simplest and most underrated action. The robots.txt file tells crawlers what they can visit on your site. Many sites, especially those built with recent CMSs, block unknown crawlers by default - and AI bots are different crawlers from Googlebot.

If ChatGPT, Claude, Perplexity and Gemini can't read your site, you're invisible to their live-search answers. Add something like this to your robots.txt:

This is the prerequisite for everything. If you skip this action, the other six matter much less.

Implement Schema.org for key entities

AI models don't "see" your site the way you do. They read it abstractly, looking for structured signals that say who you are, what you do, why you're credible. Schema.org is the shared vocabulary that lets you make this information explicit in a format AI systems immediately understand.

For most professional sites, the most important Schema.org types are:

• Person or Organization - who you are (name, role, website, social)
• Service - what you offer (description, service area, price)
• FAQPage - recurring questions and answers
• Article or BlogPosting - your content
• LocalBusiness - if you have a physical location

Schema.org doesn't guarantee citations, but it drastically reduces ambiguity: instead of hoping the model correctly infers from scattered clues, you give it a precise map. According to the LLMO/GEO/AEO framework analysis published in 2026, operators who systematize entity consistency on Schema.org report a 45% increase in AI citation frequency within two quarters.

A particularly important property is sameAs: it connects your entity to other authoritative representations (Wikipedia, LinkedIn, professional profiles). It's the main mechanism by which AI systems verify that you are really who you claim to be.

Build a consistent, repeated identity

AI systems consolidate the representation of a person or brand through consistent repetition across different sources. If you present yourself on your site as a "strategic consultant", on LinkedIn as "advisor", in an interview as "marketing specialist", the model builds a fragmented representation.

Define an official version of yourself or your company - who you are, what you do, for whom, where - and repeat it identically across all channels: site, LinkedIn, social bios, podcast guest appearances, guest articles, professional directories. Consistency matters more than creativity at this stage.

Distribute your presence on sources AI systems actually read

Your site alone isn't enough. According to Profound's 2026 analysis, in top-of-funnel queries 85% of citations in ChatGPT come from external sources, not from the subject's own site. AI systems seek corroboration: they want to see the same information about you across different, independent sources.

The sources that really matter, in order of weight:

• Wikipedia (if you meet the requirements - it's the single most cited source by ChatGPT, with nearly 48% of all citations)
• Professional directories in your sector (Crunchbase, professional bodies, trade associations, public registries)
• Third-party articles mentioning you (interviews, guest contributions, citations in industry publications)
• Cross-platform reviews (Google Business Profile, Trustpilot, TripAdvisor depending on the sector)
• Podcasts and video content with indexable transcripts

The goal isn't "link building" as it once was. It's making sure the same information about you reaches AI systems from different, independent sources.

Write citable content, not SEO content

AI engines tend to cite content with specific traits: they answer precise questions, have clear structure, include verifiable data, and themselves cite authoritative sources.

What this means in practice:

• Open every article with a direct answer to the implicit question (TL;DR)
• Use descriptive H2 and H3 headings that reflect real questions
• Include statistics with cited sources
• Reduce ambiguity: define entities, avoid vague pronouns, make every section self-contained
• Keep paragraphs short and well separated

This doesn't mean writing "for the bot". It means writing in a structured and honest way, because that's exactly what AI systems reward. Good content for AI is also good content for human readers.

Update content regularly

AI engines have two "memories": the pre-trained one (what they knew up to training cutoff) and the real-time retrieval one (what they find on the web at query time). The retrieval memory heavily rewards freshness.

Perplexity in particular gives significant weight to publication and last-update dates. But the surprising data point comes from the most recent research on schema markup: updating structured data (with the dateModified field) produces a median 22% increase in AI citations.

Practical action: review your key content (the About page, service pages, foundational articles) every 6 months. Update examples, data, links. Change the dateModified in Schema.org markup. AI systems notice.

Measure, monitor, correct

The last action, but fundamental: without measurement, you work in the dark. The representation AI systems make of you isn't static. It changes with model updates, with new content you publish, with what others write about you. It also changes over time: the same models respond slightly differently weeks apart, because they're re-trained and re-calibrated.

Without continuous monitoring, you don't know if your actions are working. You only know you're doing things, hoping.

What to monitor in concrete terms:

• What ChatGPT, Gemini and Perplexity say when you're searched
• Which sources they cite when speaking about you
• Whether there are distortions, errors, omissions in the representation
• How the representation changes month over month

Monthly monitoring is the minimum to get useful signals. Weekly is too much (models oscillate and generate noise), quarterly is too little (distortions consolidate before you notice).