How to show up in AI answers on LLMs
9 AI-specific tactics and how to implement them.
Most B2B SaaS marketers are treating generative engine optimization (GEO) as a rebrand of what they already do with SEO. However, the underlying mechanics are different.
Traditional SEO is a ranking game, where you’re competing for position on a results page that a human then chooses to click. By contrast, GEO is a citation game, where you’re competing to be the source an AI model references when it generates an answer. No click required, no position one to aim for. It simply comes down to, does the model trust your content enough to quote it.
That shift sounds subtle but it isn’t. A 2025 analysis of 300,000 keywords found that AI Overviews correlate with a 34.5% drop in average CTR for the #1 organic result compared to similar queries without them. The traffic model is breaking. If you’re still measuring success purely through rankings and click-through rates, you’re optimizing for a game that’s getting smaller.
This post covers:
What actually changes between SEO and GEO; ranking vs referencing
The 9 tactical shifts that matter, with examples of how to implement them
Where GEO and SEO genuinely overlap
An honest state of where GEO is right now
The shift from being ranked to being referenced
In traditional search, your own well-optimized pages could rank on their own merit. Write good content, build some links, structure it properly, Google notices, users click.
GEO doesn’t work that way. AI search shows a systematic bias toward earned media over brand-owned content. Much stronger than traditional search algorithms ever did. What you publish on your own domain matters less, whereas what gets written about you, by sources the model already trusts, matters enormously.
That’s the biggest strategic reorientation. Not a tactical tweak.
9 tactics that move the needle
1. Earned media is your highest-leverage investment
Getting placement in authoritative industry lists, round-ups, and third-party reviews isn’t a PR nice-to-have anymore. It’s a primary acquisition channel. When AI engines are asked to recommend tools in your category, they pull from sources they trust. If you’re not in those sources, you’re invisible.
This is the sharpest break from traditional SEO. Own-site authority still matters, but it no longer carries the conversation.
2. Write for machine scannability, not for human delight
AI search engines don’t surface the most insightful content. They surface the easiest-to-parse content. Practically, that means:
Short paragraphs of two to three lines max.
Bullet points and numbered lists wherever a sequence exists.
Lead each section with one or two sentences that directly answer the heading.
Use a consistent answer pattern: definition → detail → example.
Semrush’s research confirms this. Structure beats prose when machines are the primary consumer.
3. Target conversational queries, not typed keywords
The way someone prompts ChatGPT is different from how they’d type a Google query. Traditional keyword tools miss this entirely. “Best CRM for SMB” becomes “what CRM should a 20-person B2B company use if they’re moving off spreadsheets.”
You need to research how your audience actually phrases questions to AI assistants and then write content that answers those specific phrasings directly.
An easy way to do this research is to ask each LLM, "What questions do people ask about [your topic/product category]?" or "How would someone ask an AI assistant to help them find [your type of product/service]?" The models will surface natural language patterns they've been trained on.
Another way is to filter your Google Search Console queries for long-tail, question-based queries (anything starting with "how," "what," "why," "best way to," etc.). These are the closest proxy for AI-style queries, and they'll show you what's already driving impressions even at low volume.
4. Make your content quotable
AI-generated answers are essentially assemblies of citations. The content that gets cited is content that can be dropped in without editing; concise definitions, clean numbered steps, specific observations that stand alone.
For example, imagine you are trying to show up in AI results for the question, “what is project management software and why do teams use it?”: