Pillar guide
Generative Engine Optimization: a working guide
What GEO actually is, how AI engines pick their sources, and the playbook we run to get clients cited by ChatGPT, Perplexity, Gemini, and Google AI Overviews.
· 11 min read · by the Crescendo team
Somewhere in the last two years, your customers stopped finishing their searches. They type the question, an AI writes a paragraph, three sources get named, and that’s the whole transaction. If you’re one of the three, congratulations. If you’re not, you don’t exist for that buyer, and no rank tracker on earth will tell you it happened.
Generative Engine Optimization is the practice of becoming one of the three. This guide is the playbook we run at Crescendo, written down in enough detail to be useful whether or not you ever pay us.
What GEO actually is
GEO is the work of getting AI engines, ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews, to cite your domain or name your brand when they answer questions your buyers ask. Not “ranking” in them. There is no rank. An answer either uses you as a source or it doesn’t, and that binary is the new metric that matters. We wrote a whole piece on what counts as an AI citation because the term gets used sloppily.
The engines don’t work alike, which is the first thing that surprises people. Perplexity cites sources on every single answer. ChatGPT cites when it decides to browse, which depends on the question. AI Overviews appear on some queries and not others, and lean hard on Google’s existing index. Treating them as one channel is like treating Google and TikTok as one channel because both have a search box.
Why this got urgent
Three things converged. AI Overviews rolled out to the majority of informational queries in the US, sitting physically above the #1 organic result. ChatGPT’s weekly user count crossed into mainstream territory, and its product-recommendation answers got noticeably better sourced. And buyers, especially under 35, started going to an assistant first, using classic search only to verify.
The uncomfortable part: click-through on queries with an AI answer drops sharply, but the few clicks that remain go overwhelmingly to the cited sources. So the pie shrank and the winners’ share grew. This is a consolidation event. Most of your competitors haven’t noticed yet, which is precisely why now is the cheap time to move.
How engines choose their sources
We’ve logged thousands of citation checks across client accounts. Patterns repeat. Engines reach for sources that are:
- Unambiguous about what they are. If a model can’t state in one sentence what your company does and for whom, it won’t risk citing you. Entity clarity beats cleverness. Your homepage, About page, and third-party profiles need to agree with each other, word for word where possible.
- Directly responsive. Pages that answer the actual question in the first screenful get cited. Pages that wind up to the answer through 800 words of throat-clearing don’t, even when the answer is in there somewhere.
- Specific. Numbers, prices, dates, named comparisons. “Plans start at $79/month” gets quoted. “Affordable plans for every budget” is invisible. Engines are allergic to marketing language because it gives them nothing to synthesize.
- Recently touched. The freshness bias is real and measurable. We’ve watched pages drop out of rotation as they age and come back after an update. Details in our freshness write-up.
- Corroborated elsewhere. Engines triangulate. A claim that exists only on your domain is one source. The same claim echoed in a review site, a directory, an industry publication is a fact. This is why digital PR quietly became a GEO discipline.
The playbook
1. Find the questions that matter
Not keywords. Questions. “best crm for small agencies,” “is invisalign worth it,” “how much does it cost to sell a dental practice.” Pull them from sales calls, support tickets, and the People Also Ask boxes you already rank near. Twenty to fifty real buyer questions is enough to start. Tag them by intent: we use buy-intent, sell-intent, comparison, education, and data.
2. Establish your baseline
Run every question through every engine and log who gets cited. The result is a matrix: queries down the side, engines across the top, a check or a miss in each cell. The first baseline is always humbling. Most brands discover they’re cited on 10–20% of the questions they assumed they owned. The mechanics are in how to track AI citations.
3. Study who beat you
Every miss names the competitor who took your seat. That list, sorted by frequency, is your roadmap. Sometimes it’s a rival. More often it’s a review site or a Reddit thread, which tells you the engines couldn’t find a primary source worth trusting and settled for commentary. Full method in competitor citation analysis.
4. Build pages engines want to quote
For each priority question you’re losing, one page that answers it better than anything cited today: answer up front, real numbers, a comparison table where honest, sources for your claims, and schema that mirrors what the page says. One question, one page. Resist the 4,000-word ultimate-guide reflex; engines extract, they don’t admire.
5. Re-check, on a schedule
Citations wobble. The same question can cite you Monday and not Thursday. Single checks tell you nothing; the trend over weekly checks tells you everything. A query is “won” when it cites you in the majority of checks across a month, on the engines your buyers actually use.
What to stop doing
GEO budgets usually come from somewhere, so here’s where we take them from: high-volume informational blogging aimed at keywords nobody buys from, exact-match anchor link building, and programmatic pages that differ by one city name. All three were marginal under classic SEO. Under GEO they’re dead weight, and the thin programmatic stuff actively muddies your entity. The full comparison lives in GEO vs SEO.
Where this is headed
Our bet, and the reason we built Crescendo: assistants become the default first touch for most considered purchases, classic SERPs become the verification layer, and “citation share” becomes a line item next to “market share” in board decks. The brands that treated 2025–2026 as their land-grab window will be very hard to dislodge. Engines, like people, keep trusting who they already trust.
Common questions
- Is GEO just SEO with a new name?
- No, but they overlap heavily. Maybe 70% of solid SEO work also helps GEO. The difference is the target: SEO optimizes for a ranked list of links, GEO optimizes for being quoted inside a synthesized answer. The measurement, the content formats that win, and the competitive dynamics are different enough to need their own discipline.
- How long does GEO take to show results?
- Faster than classic SEO in our experience. Perplexity and ChatGPT can pick up a genuinely useful new page within weeks. Google AI Overviews move slower because they lean on traditional ranking signals. We tell clients to expect first citation movement inside 60 days on at least one engine.
- Can you guarantee citations?
- Nobody can, and you should run from anyone who says otherwise. The engines are non-deterministic; the same question asked twice can cite different sources. What you can do is measure citation rate over repeated checks and push it up systematically.
- Does GEO matter if my traffic is fine?
- Look at the trend, not the level. AI answers are absorbing informational queries first and commercial ones next. The brands that show up in AI answers now are building a moat that will be expensive to assault later, because engines keep citing what they've already learned to trust.
Keep reading
What is an AI citation, exactly?
AI citations are the new top of funnel. Here's what counts as one, how each engine displays them, and why they don't show up in any analytics tool you own.
GEO vs SEO: what actually changes
Most of your SEO work still matters for GEO. Some of it matters more. A few habits will actively hurt you. Here's the honest breakdown.
How to get cited in Google AI Overviews
AI Overviews now sit above the #1 result on a huge share of queries. What we've learned about earning a citation box from hundreds of live checks.
How ChatGPT picks its sources
ChatGPT answers product and vendor questions millions of times a day. Here's what its citation behavior looks like in practice and what moves the needle.