Understanding GEO is one thing. Actually showing up when your customers ask AI about your category is another. Here’s the tactical playbook.
By Matthis Duarte — Senior SEO professional, 12 years experience
If you’ve read about Generative Engine Optimization and thought “this makes sense conceptually, but what do I actually do on Monday?” — this article is for you.
Getting cited in AI-generated responses is not a single tactic. It’s a stack of compounding signals built over months. But the signals are concrete, the actions are specific, and the results are measurable. The brands earning consistent citations in ChatGPT, Perplexity, and Gemini are not doing anything mysterious — they are doing a small set of things very well and very consistently.
AI models reward clarity, specificity, structural usefulness, and external validation. They ignore vagueness, filler, and anything not grounded in verifiable evidence. Keep that filter in mind throughout everything that follows.
Understanding how each engine selects its sources
Before diving into tactics, you need to understand that ChatGPT, Perplexity, and Gemini are not the same engine. They source their information differently — which means optimisation priorities differ significantly per platform.
| Engine | Primary source | Key optimisation signal |
|---|---|---|
| ChatGPT (no browsing) | Training data (knowledge cutoff) | Long-term brand authority, entity recognition, historical web presence |
| ChatGPT (with browsing) | Real-time Bing + training data | Current search visibility + brand authority |
| Perplexity | Real-time web search | Content freshness — strong preference for pages updated within the last 30 days |
| Gemini | Google index + training data | Google ranking signals, E-E-A-T, brand-owned pages with schema, Knowledge Graph entity strength |
| Claude (Anthropic) | Training data | Entity recognition, coverage in authoritative sources |
Two takeaways worth internalising: Perplexity is the most immediately responsive to fresh content — update a page today and it can appear in Perplexity citations within days. Gemini has a notable bias toward brand-owned websites with proper schema and consistent subdomains. ChatGPT’s training-data dependency makes it the longest game — but entity signals you build today will compound over time.
Step 1 — Become a clearly defined entity
This is the foundation. LLMs reason about the world through named entities — companies, people, products, concepts — and the relationships between them. A brand that exists only on its own website, described only in its own words, is hard for a language model to reference confidently.
Your goal: make your brand a well-documented entity across multiple authoritative, independent sources.
Wikipedia. If your brand or founder meets Wikipedia’s notability criteria, a Wikipedia article is one of the highest-value GEO signals available. Wikipedia is heavily represented in LLM training data and creates a canonical, third-party description that models lean on.
Wikidata. Even without a Wikipedia article, a Wikidata entry establishes machine-readable entity data — brand name, category, founding date, key people, relationships. This is directly consumed by knowledge graph systems including Google’s.
Crunchbase, LinkedIn, and industry directories. The name, category, description, and key people should be identical across all platforms. Consistency of entity signals across sources is what gives models confidence to cite you accurately.
Google Knowledge Panel. For Gemini specifically, your presence in Google’s Knowledge Graph matters enormously. Claim your Knowledge Panel if eligible and ensure entity data is consistent with what Google expects.
Step 2 — Own a topic, not just a brand name
Being known as a brand is not enough. You need to be the entity most strongly associated with specific topics your customers ask AI about.
When a user asks Perplexity “what’s the best approach to topical authority for a SaaS startup?”, the engine doesn’t search for brand names — it generates an answer drawing on its assessment of which sources have consistently covered that topic with authority. Your goal is to become the default citation for a defined set of questions.
Choose 3–5 core topics you want to own. These should map directly to the questions your ideal customers are asking AI most frequently. For a growth marketing consultancy: growth hacking frameworks, startup SEO, AI visibility strategy, product-led growth, SaaS content strategy.
Publish consistently, not just heavily. A brand that has published 30 high-quality articles on startup SEO over 18 months is more likely to be cited than one that published 10 excellent articles last month. AI systems weight consistency over volume.
Update existing content regularly. For Perplexity in particular — which favours pages updated within the last 30 days — a refresh of a key article can meaningfully improve citation frequency. Recency is a ranking signal.
Step 3 — Earn authoritative third-party mentions
AI citations don’t come from your website alone. They come from the pattern of who cites you across the broader web. This creates a compounding effect: brands mentioned in authoritative industry publications accumulate citation authority that makes them more likely to appear in AI responses — a concept sometimes called earned mentions.
Target industry publications with genuine editorial standards. One mention in TechCrunch, VentureBeat, or Search Engine Journal is worth more for AI citation authority than hundreds of mentions on low-authority blogs. The quality of the source matters enormously.
Pursue podcast appearances and interviews. Podcast transcripts and show notes are indexed and scraped. Consistent appearances across industry podcasts build a clear third-party association between your brand and your topic area.
Publish original research. Data studies, surveys, and proprietary benchmarks are the highest-leverage GEO content play available. When your data is cited by other publications, your brand travels with that citation into AI training data and real-time search results. The research becomes the citation anchor.
🔴 Case study — Ahrefs: topic ownership at scale
Ahrefs built its AI visibility without any GEO-specific strategy — simply through a decade of publishing the most comprehensive SEO content on the internet. Every SEO topic covered in exhaustive depth, consistently updated, from a domain with exceptional authority.
The result: when users ask any AI engine about SEO — keyword research, backlink analysis, content strategy, technical SEO — Ahrefs appears disproportionately. Not because it optimised for AI citation, but because it did the foundational work — topic ownership, third-party authority, entity clarity — at a scale that naturally translated into AI visibility.
→ Result: Ahrefs became one of the most-cited SEO brands in AI-generated responses, built on a content strategy that predates GEO as a concept by years.
Step 4 — Structure your content for extraction
AI summarisers pull individual passages from pages — they don’t read articles holistically. Content that is easy to extract gets cited more. This means formatting is not just a UX decision — it’s a citation signal.
Structure your most important content as:
- Step-by-step guides — numbered, self-contained, easy to lift as a sequence
- Comparison tables — AI loves presenting structured comparisons
- Direct Q&A blocks — question as a heading, direct answer in the first sentence
- Concise definitions — “[Term] is [definition].” — the pattern AI models extract most reliably
Back everything with specific data, named examples, and verifiable claims. AI models are trained to favour content that enhances the practical value of a generated response. Vague overviews don’t get cited. Specific, actionable answers do.
Schema markup for AI readability: FAQPage schema, HowTo schema, Article schema with explicit author markup, and Organisation schema all help AI crawlers understand what your content is about and who stands behind it — affecting how confidently models reference you.
Step 5 — Measure your AI visibility
You cannot optimise what you don’t measure. AI visibility measurement is still early-stage, but practical approaches exist now.
Manual prompt testing. Build a set of 20–30 prompts reflecting the questions your customers ask AI about your category. Run them weekly across ChatGPT, Perplexity, and Gemini. Track: Is your brand mentioned? How is it described? Which competitors appear instead?
Dedicated AI monitoring tools. Platforms like Profound are building AI citation tracking specifically for brand monitoring in LLM responses — the GEO equivalent of rank trackers. [verify]
Referral traffic from AI engines. Both Perplexity and Gemini generate referral traffic when they cite sources. Monitor referral traffic from perplexity.ai in your analytics — a spike is a direct signal your content is being cited.
Entity recognition test. Ask ChatGPT: “What do you know about [your brand]?” The accuracy and confidence of the response tells you how well your brand is represented in training data — and where the gaps are.
Key takeaways
- ✓ Each engine works differently: Perplexity favours content updated within 30 days; Gemini favours brand-owned pages with schema; ChatGPT relies heavily on training data and entity recognition — optimise accordingly
- ✓ Entity clarity is the foundation — your brand must be consistently described across Wikipedia, Wikidata, Crunchbase, and industry directories before anything else compounds
- ✓ Topic ownership beats brand awareness — become the most cited source for 3–5 specific questions, not just a recognisable name
- ✓ Earned mentions from authoritative publications are the highest-leverage citation signal — one TechCrunch mention outweighs hundreds of low-authority references
- ✓ Original research is the best GEO content play — your data becomes the citation anchor and your brand travels with it into AI responses across the web
- ✓ Measure weekly with manual prompt testing and referral traffic monitoring — it’s the only way to know if the strategy is working before the tools fully catch up
Matthis Duarte is a senior SEO professional with 12 years of experience. HackingStory.com reverse-engineers how the fastest-growing startups actually grew — with real data, not press releases.
