GEO and User Context: Why AI Recommends Your Competitor to One User and You to Another

AI Doesn't Recommend the Same Sources to Every User, and That's Intentional

When two users ask an identical question to ChatGPT, Perplexity, or Gemini, they often receive answers citing different sources. This variation isn't a malfunction, but the result of an implicit profiling system that AI engines build silently around each conversation. AI analyzes who you are (your industry, your location, your presumed expertise, your search history), then adapts not only the tone and format of the response, but also the sources it cites as credible. This means optimizing for "the first position on an AI" is an outdated strategy: you need to optimize for a specific user profile, not for a generic query.

How AI Builds an Implicit User Profile

AI answer engines don't start from scratch. From the very first interaction, several signals contribute to building a silent user profile:

The Role of Cookies and Session Tokens

If a user is logged into ChatGPT Plus, a Google account, or a Perplexity profile, the AI has explicit information: conversation history, saved preferences, declared user profile. But even an unauthenticated user leaves traces. Session tokens, persistent cookies, and IP addresses allow AI to recognize a returning user and retain contextual memory, even implicitly. This persistence means that an initial question like "how to start a blog" can be reinterpreted in light of a previous query about "enterprise content management tools."

Passive Detection of Industry and Use Case (B2B vs B2C)

AI analyzes vocabulary, the specificity of questions, and conversational context to infer whether you operate in B2B or B2C. A user who asks "how to integrate a payment API into a SaaS platform" will be treated differently from one who asks "how to accept online payments." The AI implicitly detects that the first belongs to a technical professional context, the second to a personal business. It will then recommend sources relevant to each segment: cutting-edge technical resources for the first, accessible tutorials for the second. The same domain (for example, a blog about online payments) can be cited or ignored depending on this inferred profile.

Language and Geographic Region

IP address, interface language, browser location settings, and question content (references to local laws, currencies, regional providers) allow AI to locate the user. A question about the "best CRM tools" asked from Switzerland won't receive the same recommendations as the same question asked from Canada. AI favors locally relevant sources, regionally established providers, and content written according to the region's norms and legal context. A US-based site can be widely cited for an American audience, but ignored for a user in France if the content doesn't address GDPR regulations or French specifics.

Presumed Level of Expertise

From the first questions, AI infers the user's level of expertise. A user who uses cutting-edge technical terms ("tokenization," "fine-tuning," "inference latency") will be classified as an expert; one who asks basic questions will be treated as a beginner. This classification directly influences the level of complexity and the sources recommended. An expert will receive links to academic papers, deep technical documentation, and content written for seasoned practitioners. A beginner will receive introductory guides, progressive tutorials, and simplified explanations. The same site can be ignored as "too basic" for one segment and recommended as "perfect for getting started" for another.

How Perplexity, ChatGPT, and Gemini Build and Use These Profiles

Each AI answer engine uses slightly different mechanisms, but the principle remains the same: adapt citations based on the detected profile.

Perplexity and Profiling Through Custom Copilots

Perplexity allows logged-in users to create custom "copilots," defined by instructions, preferred sources, and search parameters. But beyond this explicit personalization, Perplexity silently analyzes search patterns: domains consulted, depth of questions, topic history. A user who consistently searches on marketing topics gets responses calibrated for a marketer; one who searches on cloud infrastructure receives responses calibrated for a DevOps engineer. The sources cited change accordingly: a reputable marketing blog will be cited for the first, a technical article will eclipse the same blog for the second.

ChatGPT and Conversational Profile Persistence

ChatGPT builds a progressive profile across each conversation. OpenAI indexes conversation histories (for authenticated users) and analyzes them to refine an implicit model of who you are. But equally important: within a single conversation, each response refines the profile. If you ask a basic question, then a second very technical one on the same topic, ChatGPT will recalibrate its expertise profile for the third question. This fluidity means that the position of sources cited in responses changes dynamically based on the trajectory of the conversation, not based on static segmentation. A site can be ignored in a general response, but cited in a follow-up response if the user has shown growing expertise.

Gemini and Profiling Through the Google Ecosystem

Gemini (formerly Bard) has a unique advantage: access to the complete Google ecosystem. If a user is authenticated via a Google account, Gemini has signals from Google Analytics, Google search history, Google Workspace apps used, and even Gmail emails. This ecosystem context creates an extraordinarily precise user profile. A user who uses Google Workspace Pro at an agency receives recommendations oriented toward enterprise solutions; one who uses Gmail personally receives consumer content. The same sources are weighted differently depending on this implicit ecosystem context.

The Signals AI Uses to Adapt Citations

Beyond user profiling, AI also evaluates each candidate source based on how it matches the detected user's context:

Publication Date and Content Freshness

A user searching for emerging technologies receives responses based on recent content; one searching for fundamentals may receive older but canonical references. AI weights freshness according to the domain and context: a 2024 SEO article will be preferred for an evolving topic (AI technologies, Google algorithms), while a 2015 classic will remain relevant for a stable topic (basic marketing principles).

Site Localization and Regional Relevance

Content published on a France-based site, in French, addressing GDPR regulations will be highly valued for a French user; the same content will be devalued for an American user. AI detects the site's base country (via domain, infrastructure, mentioned legal context) and matches it with the user's location. This explains why two users don't see the same sources for an apparently identical question: one is optimized for their region, the other for theirs.

Domain Authority, Weighted by Industry

A site's authority isn't a universal metric. A highly authoritative site in finance may carry little weight for a UX design question. AI weights domain authority based on the user's detected industry and the topic of the question. It implicitly analyzes: is this domain a recognized authority in this specific industry? For this user? With this level of expertise? A popular generalist site can cite niche experts for an expert user, and clear generalists for a beginner.

Ideological and Contextual Alignment

AI silently analyzes potential ideological biases and the contextual alignment of content. A user searching for the "digital transformation of an SME" will receive sources aligned with SME realities; one searching for the same question in an enterprise context will receive enterprise-oriented sources. This isn't political censorship, but contextual adaptation: choosing sources relevant to the specific detected context.

Why Your Competitor Is Recommended to One User and You to Another

Now, the mechanics become clear. Suppose you and your competitor both sell project management software. A B2C user searching for "simple collaborative tool" may receive an answer citing your competitor, who has built a reputation as a "consumer-friendly" solution. The same tool for a B2B tech user receives an answer citing your site, because you've positioned yourself as a "solution for technical teams." The AI doesn't like your competitor better in absolute terms; it adapts the citation based on who's asking the question.

Similarly, if your content is written for French GDPR compliance experts and an American user searches for "how to structure a data team," the AI probably won't cite your site, even if excellent, because it addresses the wrong segment. Your competitor, who wrote a guide on "data team structure for US startups," will receive the citation.

This adaptation also explains why a traditional SEO strategy (optimizing for "the first position") fails in an AI context: there isn't one first position, there are first positions per segment. Your competitor may occupy the first position for one segment, you for another, and a third party for a third, all on the same query.

How to Optimize Your Visibility with Specific User Segments

Explicitly Define Your Target Segment and Write for It

Choose a precise user segment: its industry (B2B SaaS, e-commerce, finance), its geography (France, Belgium, French-speaking Switzerland), its level of expertise (beginner, intermediate, expert), its use case (growing startup, established SME, enterprise). Then, write all your content as if it were addressed only to that segment. Don't aim for "everyone" with generic content. Write guides that explicitly address the specific challenges of your segment, using their vocabulary, referencing their legal and regulatory contexts, mentioning the tools and solutions they already use.

When you write "how to implement an online payment system for a French e-commerce business," AI implicitly detects that your content targets a specific segment (French e-commerce), and it recommends you massively for users matching that profile. When you write "online payment guide," you compete with everyone and dominate no one.

Integrate Localization and Regional Context Signals

Explicitly mention your target region or language in your content. Reference applicable laws (GDPR for Europe, CCPA for California), local currencies, regionally established providers, regional accounting standards. Publish on a domain with an appropriate geographic extension (for example: .fr for France) or, if you use a .com, mention your physical and legal bases in your content and metadata.

AI heavily weights explicit localization signals. A .de site written in German addressing German law will be strongly recommended for German users; a .com site addressing the same audience without these signals will be devalued for that segment, even if its content is excellent.

Structure Content for Distinct Levels of Expertise

Instead of writing a single guide "from beginner to expert," write three separate pieces of content: a guide "for beginners," one "for intermediates," one "for experts." AI implicitly analyzes the level of complexity of content and matches it with the user's detected expertise profile. When a beginner searches, they receive the beginner guide; when an expert searches, they receive the expert guide. Three distinct pieces of content give you three chances to be cited, one per expertise segment.

In each piece of content, use vocabulary consistent with the level of expertise. The beginner guide explains terms; the expert guide assumes they're known and moves faster. AI detects this consistency and aligns the citation with the detected context.

Clarify Your Industry Positioning (B2B vs B2C, Specific Industry)

Be explicit about your target industry in your content and positioning. If you address B2B SMEs, write case studies about SMEs, reference SME challenges, mention typical SME budgets and resources. AI detects this positioning implicitly and recommends your content when a B2B SME user searches.

Similarly, if you address a specific industry (healthcare, finance, retail, manufacturing), mention it explicitly, describe the specific challenges of that industry, reference sector-specific regulations. You then become dominant for users in that industry, even if you're less known globally.

Optimize Metadata for User Context

Your title tags, meta descriptions, and structured schemas must reflect the target segment of each piece of content. A page addressing French DevOps engineers must signal this in its metadata. A page addressing growing startups must mention "startup" or "growth stage" in its metadata. AI reads these signals implicitly and uses this metadata to refine the matching between content and the detected user profile.

Create Content That Addresses User Transformations and Trajectories

AI also adapts its citations based on the user's implicit trajectory. Someone who first asks beginner questions, then advanced questions, is treated as someone growing. Create content that recognizes and addresses these trajectories: "Guide for SMEs Ready to Scale," "How to Move from Freelancer to Agency," "How to Move Your On-Premise Infrastructure to the Cloud."

These pieces of content target a specific user profile in terms of trajectory, not just current status. They become dominant for users in this transition phase, because AI detects them as aligned with the detected evolving context.

Publish in Languages and Regional Variants

If you serve multiple linguistic regions, don't translate passively. Publish distinct versions, written natively for each region. Quebec French differs from France French; Indian English differs from American English. AI detects these variants implicitly and recommends content written natively for the user's region.

A site that publishes in generic "French" will be recommended less specifically than a site publishing distinctly in "France French" and "Belgian French." This writing distinction signals to AI a deep regional knowledge and commitment.

Measure and Validate Your Visibility with Specific Segments

Unlike traditional SEO where you track your ranking for a given query, optimizing for AI recommendation by segment requires different measurement methods.

Test Manually with Varied User Profiles

Create user profiles representing your target segments, then ask questions through these profiles: from a French VPN location, from another located in the United States, from a new user account, from an account with history on B2B vs B2C topics. Observe which sources AI recommends. You'll quickly see if your content is recommended for the target segment and ignored for others. If so, your strategy is working. If not, your content doesn't explicitly address your segment enough.

Analyze Usage and Traffic Patterns by Segment

In your analytics, segment traffic by source, geolocation, device type, session duration. Do users arriving from Perplexity in French from France behave differently from those arriving from ChatGPT in English from the United States? If your content effectively targets a segment, you'll see traffic concentrated on that segment, high engagement (pages per session, duration), and a low bounce rate. Other segments arrive little or bounce quickly. This is the sign that you're correctly visible for your segment and that your content satisfies it.

Use Specific AI Monitoring Tools

Tools like Semrush, Ahrefs, and SEMrush are starting to offer AI recommendation tracking. Specialized solutions are also emerging. Use them to monitor whether your site is cited by Perplexity, ChatGPT, and Gemini for queries in your domain. But don't settle for tracking "your site is cited" globally; segment by query, by detected user type (if possible), by location. You'll see where your visibility is strong (high citation rate) and where it's weak.

Pitfalls to Avoid

Don't Assume Traditional SEO Optimization Is Enough

Google values domain authority and external links. AI also values contextual alignment with the detected profile. A site that is highly authoritative globally but not aligned with the segment can be ignored for that segment, even if it dominates Google. Conversely, a smaller site but extremely aligned with a segment can be recommended massively by AI for that segment, even if it doesn't rank well on Google in general.

Don't Aim for "the First Position" or "Being Cited in AI Responses"

These goals are too generic. You can't be "the first position" for a query if different segments receive different answers. Instead, aim for: "being the number 1 recommended source for French tech startups looking for CI/CD tools." That's a specific, measurable, and realistic goal.

Don't Ignore Localization and Regional Relevance

Even if your product or service is global, your content must be regionally specific. AI heavily weights regional signals. Ignoring localization means letting your regional competitors dominate with regional users, even if you're better in absolute terms.

Don't Over-Diversify Your Positioning

If you write for "everyone" (beginners and experts, B2B and B2C, multiple regions), you dominate no segment. Better to dominate one segment than to be visible for all segments without dominating any. Specialize initially, then diversify progressively.

Conclusion: From Ranking to Contextual Recommendation

The era of unified search engine rankings is giving way to an era of contextual recommendations by AI. Your competitor is recommended to one user and you to another not because the AI likes them better in absolute terms, but because their respective content is better aligned with the different detected profiles. Winning at AI visibility requires accepting this fragmentation: it's not a single position you occupy, but a set of different positions with different user segments. Define your target segments precisely, write for them with intentionality, and AI will recommend you massively to them. Try to serve everyone and you'll dominate no one.