GEO in Year One: A Realistic Roadmap From Zero to Citable in ,0,6
Most brands starting GEO in ,0,6 will see their first AI mentions within a few weeks. Getting cited consistently across ChatGPT, Perplexity, Claude, and Google AI Overviews takes longer, usually three to six months of deliberate work. This roadmap covers what to do, in what order, and why the sequence matters.
The honest version of year one in GEO looks nothing like the hype. There's no single fix that makes ChatGPT start recommending you overnight. What there is: a set of well-ordered actions that compound. Get the foundations wrong early and the rest of the year is wasted. Get them right and by month six you'll have measurable visibility where your competitors have none.
The order in which you do things in GEO determines whether your later efforts have any effect. If you spend months producing content before your site is crawlable by AI bots, none of that content enters the retrieval pool. If you track prompts before you've defined a proper prompt set, you'll optimise for the wrong queries for months.
GEO has three distinct phases in year one: technical access (can AI engines see you?), content authority (do AI engines have a reason to cite you?), and earned visibility (do third-party sources mention you in ways AI engines trust?). These phases overlap but they don't collapse. Technical access has to come first.
Before anything else, confirm that AI crawlers can access your site. This takes a few hours and unlocks everything else. Most brands skip this step and wonder why their content never appears in citations.
Check your robots.txt file. Several AI crawlers have their own user-agent strings and will be blocked by wildcard disallow rules. You need to explicitly allow the crawlers that matter:
- OAI-SearchBot and ChatGPT-User (OpenAI, used by ChatGPT Search via Bing)
- PerplexityBot (Perplexity's own crawler)
- ClaudeBot and anthropic-ai (Anthropic, used by Claude via Brave Search)
- Google-Extended (used by Gemini and Google AI Overviews)
Beyond robots.txt, confirm that your key pages render without JavaScript. AI crawlers are generally less capable of executing JavaScript than a modern browser. If your content lives inside a React or Vue app that renders client-side only, crawlers may fetch an empty page. Server-side rendering or static HTML is the safer architecture.
Also check Bing's index directly. ChatGPT Search retrieves live web content via Bing. If Bing hasn't indexed a page, ChatGPT can't cite it in real-time retrieval. Use Bing Webmaster Tools to confirm coverage and flag any crawl errors.
Claude uses Brave Search as its retrieval index. Check that your most important pages appear in Brave's results for your brand and category terms. If they don't, Brave has a URL submission tool in its webmaster portal.
You cannot measure GEO progress without a structured prompt set. This is the step that most teams either skip or do badly, and it undermines everything that follows.
A prompt set is the collection of queries you'll submit to AI engines regularly to measure whether your brand appears. The mistake most brands make here is tracking only branded queries: "[Brand] vs [Competitor]" or "[Brand] review." Those queries test competitive positioning. They don't tell you whether you're visible during the discovery phase, when a buyer asks "what's the best [category] for [use case]?" and your brand's name hasn't entered the conversation yet.
A well-structured prompt set spans six intent types. Category prompts test baseline awareness. Use-case prompts test contextual relevance. Comparison prompts test competitive positioning. Recommendation prompts test whether AI engines suggest you unprompted. Problem-solution prompts test whether you appear when someone describes a problem your product solves. Feature-specific prompts test whether you're associated with the capabilities you claim to have.
Building this set from scratch takes real search data, not guesswork. You need keyword volume data, People Also Ask patterns, and competitor analysis to understand how your category is actually queried. At BrandPrompts, we built our entire product around this problem because we kept watching brands track 15 branded queries and call it GEO monitoring. It isn't. The discovery queries are where AI visibility is won or lost, and those require genuine research to find.
For statistical reliability, aim for at least 30-50 prompts per topic-market combination. Fewer than that and you're measuring noise.
AI engines retrieve content in chunks, not full pages. A model pulling an answer from your site doesn't read the whole article. It extracts the most relevant passage. This changes how you should write.
Every section of every page needs to open with a self-contained answer in the first two or three sentences. The supporting detail follows. If the opening of a section only makes sense when read after everything before it, the AI can't use it. Write as if every paragraph might be the only paragraph someone reads.
Heading hierarchy matters more in GEO than it ever did in SEO. A strict H1 to H, to H3 structure tells AI engines how your content is organised. Question-format H,s help because they mirror how users actually prompt. "How does [Product] handle [Feature]?" as a heading is more likely to be matched to that exact query than "Feature Overview" as a heading.
Lists and tables get cited at far higher rates than buried prose. Anything comparable belongs in a table. Any set of items belongs in a list. This isn't a style preference. It reflects how AI retrieval systems chunk and score content.
Include a recency signal in your H1 and first paragraph. The year, or a phrase like "updated for ,0,6," signals to retrieval systems that your content is current. AI-cited pages skew heavily toward recently published or updated content.
This is the part of GEO that trips up most brands with strong traditional content marketing. Thought leadership that restates what's already well-known gets ignored. AI engines can generate that content themselves from training data. They have no reason to cite you for it.
What gets cited is what only you can say. That means original data from your product, anonymised benchmarks from your customer base, proprietary research, and direct experience with specific tools and outcomes. It means category definitions that establish your point of view rather than summarising the consensus. It means contrarian positions that are defensible and backed by real reasoning.
| Content Type | Likely to Be Cited? | Why |
|---|---|---|
| Original research with proprietary data | Yes | AI can't reproduce it from training data |
| Detailed how-to guides with specific steps | Yes | Structured, extractable, useful |
| Comparison tables with real specs | Yes | Structured format, high information density |
| FAQ blocks with direct answers | Yes | Mirrors query format; extractable pairs |
| Thought leadership restating known positions | No | AI can generate this from training data |
| Generic product descriptions | No | Low information density, not unique |
| Content behind logins or paywalls | No | Crawlers can't access it |
Across all major AI engines, the large majority of citations come from third-party sources, not brand-owned pages. Gemini cites comparatively more brand-owned content than its peers, but even there, earned media dominates. Claude and ChatGPT skew most heavily toward third-party sources.
This is the most important structural reality of GEO, and it's the one most brands resist because earned media is slower and less controllable than publishing your own content. But the data is consistent: AI engines trust what others say about you far more than what you say about yourself.
The practical implication is that a significant portion of your year-one GEO effort should be off-page. That means getting into industry round-ups, review platforms like G, and Capterra, and editorial coverage in publications your category considers authoritative. It means engaging genuinely on Reddit in subreddits where your buyers ask questions. It means being mentioned alongside your competitors in comparison articles written by third parties.
Reddit is a meaningful citation source for both ChatGPT and Perplexity. Not gaming Reddit, not promotional posting, but answering real questions in category-relevant communities with specific, useful detail. That content gets indexed and retrieved.
Wikipedia presence, where achievable, is high-use. The encyclopaedia is disproportionately represented in LLM training data. A Wikidata entry for your organisation, or a mention in a relevant Wikipedia category article, carries significant weight. For most brands this isn't achievable in year one, but it's worth tracking as a long-term goal.
By month six, you should have baseline visibility data across at least four AI engines. ChatGPT, Perplexity, Claude, and Google AI Overviews are the minimum set. Gemini is worth adding if your category has strong B,C or Google-ecosystem exposure.
Visibility varies substantially between platforms. A brand that appears consistently in Perplexity answers may barely register in Claude's responses. This happens because each engine has different retrieval mechanisms, different training data weights, and different source preferences. Optimising for one platform doesn't guarantee visibility on another. Platform-specific work is necessary.
For ChatGPT, Bing indexing and Wikipedia-style authority matter most. For Claude, Brave Search indexing and recency signals in content. For Perplexity, plain language, community platform presence, and accessible content without paywalls. For Google AI Overviews, traditional SEO quality remains the strongest predictor of inclusion. The platforms aren't identical problems.
Use the second half of the year to identify which prompt types your brand is winning and which it's absent from. If you're visible in comparison prompts but invisible in use-case prompts, your content strategy has a gap. If you're cited on ChatGPT but absent from Perplexity, your off-page presence in the sources Perplexity favours (Reddit, news, editorial) needs work.
Structured prompt tracking through a platform like Peec AI, Profound, or Searchable gives you this data systematically. If you want to ensure you're tracking the right prompts in the first place, see how BrandPrompts approaches prompt set construction before you set up your monitoring.
How long does it take to get first mentioned in AI search?
For a brand with existing domain authority and content, first mentions in AI search can appear within a few weeks of making content accessible to AI crawlers. For newer brands starting from a low authority baseline, expect two to four months before consistent mentions begin. The timeline depends heavily on how competitive your category is and how much earned media you can generate.
Do I need to be on all AI platforms from day one?
No. Start with ChatGPT and Perplexity. They have the largest user bases in most B1 billion and B,C categories and will give you the fastest feedback on whether your strategy is working. Add Claude and Google AI Overviews by month three. Gemini is worth adding later, particularly if your audience is heavily in the Google ecosystem.
How many prompts do I need to track?
For reliable visibility measurement, you need at least 30-50 prompts per topic-market combination. Tracking fewer means random variation in AI responses will make your data misleading. Most brands underestimate how many prompts they need because they focus only on branded queries and miss the category and use-case queries where discovery actually happens.
Does traditional SEO still help with GEO?
Yes, particularly for Google AI Overviews. A strong correlation exists between top organic rankings and AI Overview citation for that platform. For ChatGPT, Bing ranking is the more direct lever. SEO and GEO share many technical foundations: crawlability, structured content, authoritative external links. They're not the same discipline, but good SEO doesn't work against GEO.
What's the biggest mistake brands make in year one?
Over-indexing on branded prompts and brand-owned content. Brands track whether ChatGPT mentions them when someone searches their name, and they publish content on their own blog. Both matter, but neither is where GEO is won. The real opportunity is category and use-case queries where your brand name doesn't appear in the prompt. Those are the moments when a buyer is forming a shortlist and you either get on it or you don't. Most brands don't track those queries at all.
Track your brand's AI search visibility
BrandPrompts monitors how your brand appears across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Know where you stand before your competitors do.
Get started freeOr calculate how many prompts you need to track →