Something shifted quietly in how people find information, and I think a lot of marketing teams haven’t fully registered it yet. People are no longer just Googling. They’re asking ChatGPT which agency to hire, prompting Perplexity to compare software tools, and letting Gemini summarise their options before they visit a single website.
The numbers make this hard to ignore. AI chatbot sessions now reach 1.2 billion per month globally, and Gartner projects that traditional search engine query volume will fall by 25% by 2026 as answer engines continue to gain ground. More telling still: AI search traffic converts at 14.2%, compared to Google’s 2.8%. The people arriving from AI citations are already further along in their decision-making.
At WeAreBrain, we’ve been thinking hard about what this means for brand visibility. This article is our honest breakdown of LLM optimisation: what it is, how it works, and what you can do about it right now.
What is LLM optimisation?
LLM optimisation (LLMO), sometimes called Generative Engine Optimisation (GEO) or Answer Engine Optimisation (AEO), is the practice of making your brand one of the sources an AI assistant cites when it answers a relevant question.
It’s worth being precise about these terms, because they’re used interchangeably but have slightly different emphases:
| Term | What it targets | Core mechanic |
| LLMO | Any large language model platform | Citation and brand recall across all AI surfaces |
| GEO | Generative answer engines (ChatGPT, Perplexity, Gemini) | Content structure for AI-generated responses |
| AEO | Featured snippets and direct answers in traditional search | Formatting for zero-click search results |
The distinction from traditional SEO matters. SEO optimises for ranking position in a list of links. LLMO optimises for citation within an AI-generated answer. And while Google serves ten links per query, most AI assistants cite somewhere between two and seven sources. The shortlist is much shorter, which makes getting onto it both more valuable and more competitive.
Princeton University research, published at ACM SIGKDD 2024, found that specific content techniques can boost AI visibility by up to 40%. The highest-impact strategies were adding statistics with source attribution, writing in authoritative and specific language, and citing credible organisations within the content itself.
How AI assistants actually retrieve information
Understanding the retrieval process is the most useful thing you can do before trying to optimise for it. AI assistants don’t search the web the same way Google does. Most combine three distinct information sources to generate their answers:
| Source | What it is | How it works |
| Original index | The model’s training data | A vast web crawl captured months or years before deployment. Your brand’s presence here gives AI a baseline of familiarity before any query is made. |
| Review and list sources | Third-party directories, rankings, and review platforms | When asked for recommendations, AI assistants actively pull from G2, Trustpilot, Capterra, industry directories, and curated lists, often prioritising these over brand-owned pages. |
| Web search | Live retrieval at query time | The assistant searches the web in real time, reads the top pages, and blends what it finds with what it already knows to synthesise a current answer. |
The critical insight is that there is no single best route to appearing in AI answers. Brands that focus only on their own website content, or only on SEO, tend to underperform because AI systems weight all three sources simultaneously. The most consistently cited brands show up across all three: they have a credible training data footprint, strong third-party mentions, and well-structured content that live retrieval can extract quickly. A combined approach is the only reliable one.
How AI search differs from traditional SEO
The optimisation logic is genuinely different, and conflating the two leads to wasted effort. Here’s how the two systems compare:
| Factor | Traditional SEO | AI citation (LLMO/GEO) |
| Goal | Rank in a list of links | Be cited in a synthesised answer |
| Key signal | Backlinks, keyword relevance | Factual density, authority, entity trust |
| Content format | Long-form, keyword-rich | Answer-first, specific, quotable |
| Source volume | 10 results per query | 2–7 citations per answer |
| Traffic type | Click-based | Often zero-click; higher intent on arrival |
| Update frequency | Ranks shift gradually | AI answers can change query-to-query |
One pattern that stood out in recent B2B AI referral data is that the market has already fragmented significantly. ChatGPT, Claude, Gemini, and Perplexity now together account for nearly all measurable AI referrals, each with distinct retrieval logic. Optimising only for ChatGPT, as many teams did through 2024 and early 2025, now covers considerably less of the landscape than it did a year ago.
What increases your chances of being cited
This is the practical core, and the good news is that most of it is achievable without a complete content overhaul.
- Lead with the answer. Put your direct answer in the first 30% of the page. Research suggests this is where the majority of AI citations are drawn from. State what you know, then explain it.
- Write specifically and quotably. Vague marketing language doesn’t get cited. Specific claims, named statistics, and dated evidence do. Definitive language is cited roughly twice as often as hedged generalities, according to the Princeton GEO research. If you can replace “we help companies grow” with “we’ve helped B2B SaaS companies reduce their sales cycle by an average of six weeks,” do it.
- Use structured headings and lists. Short paragraphs, descriptive H2s, and bullet lists are easy for AI systems to extract and use verbatim. Dense prose is harder to parse at speed.
- Cite credible sources within your content. Linking to studies, official data, and reputable publications signals that your page itself is trustworthy, which carries weight in citation selection.
- Keep content fresh. The majority of AI crawler traffic targets content published within the past year. Build a refresh schedule for your highest-value pages, especially service pages and cornerstone articles.
- Maintain entity consistency. Your company name, description, and core details should be identical across your website, LinkedIn, Google Business Profile, Crunchbase, and any directory listings. Inconsistency creates ambiguity for AI systems trying to build a coherent picture of what your brand is and does.
- Add schema markup. FAQ, Article, and Product schema help both Google and AI systems understand the structure of your page. It’s one of the higher-leverage technical changes available, and most CMS platforms support it without custom development.
- Check your robots.txt. Many companies have inadvertently blocked AI crawlers without realising it. To allow the major ones, your robots.txt should explicitly permit GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, and Google-Extended.
- Add an llms.txt file. An llms.txt file is a plain-text document placed at your domain root that gives AI models a structured summary of what your site contains and how it should be used. Think of it as a briefing document for AI crawlers, distinct from robots.txt which controls access. It’s an emerging standard, not yet universal, but early adoption is worth the modest effort.
Why third-party mentions matter more than your own site
One of the more counterintuitive things about AI citation behaviour is how heavily AI assistants weight external sources over your own content. Companies with active G2, Capterra, or Trustpilot profiles are approximately three times more likely to be cited than those without them. Brands mentioned on Reddit or Quora show citation rates around four times higher. And presence across four or more trusted platforms correlates with nearly three times greater AI visibility overall.
We saw this pattern clearly with a B2B consulting client we worked with last year. Their website content was strong, but their external footprint was thin: no review platform presence, minimal community participation, and an inconsistent company description across directories. Once they addressed those gaps systematically, their AI citation rate across Perplexity and ChatGPT’s browsing mode improved meaningfully within a quarter.
Where to focus first:
- Review sites: G2, Capterra, Trustpilot, Glassdoor, whichever fits your sector
- Listings: Google Business Profile, LinkedIn Company, Crunchbase, Bing Places
- Communities: Reddit, Quora, Stack Exchange, answer questions in your domain genuinely
- Wikipedia and Wikidata: if editorially appropriate, LLMs draw heavily from both
Common mistakes to avoid
Most brands new to LLMO make the same errors:
- Blocking AI crawlers by default in robots.txt, cutting off live retrieval entirely
- Publishing vague copy full of positioning language and short on specific, citable claims
- Treating LLMO like keyword stuffing, optimising for search terms rather than genuine authority signals
- Inconsistent entity data across platforms, making it harder for AI systems to resolve your brand confidently
- Ignoring third-party presence while investing exclusively in owned channels
A 90-day LLM optimisation checklist
Most of this is achievable with existing resources. The point isn’t to do everything at once, it’s to build the signal over time.
| Timeframe | Action |
| Weeks 1–4 | Audit robots.txt and allow major AI crawlers |
| Weeks 1–4 | Add llms.txt to your domain root |
| Weeks 1–4 | Complete or update G2, Trustpilot, and Google Business Profile |
| Weeks 1–4 | Align your company description consistently across all platforms |
| Weeks 5–8 | Add FAQ and Article schema markup to key pages |
| Weeks 5–8 | Rewrite top service and landing pages to lead with direct answers |
| Weeks 5–8 | Add specific statistics and cited evidence to cornerstone content |
| Weeks 5–8 | Begin genuine participation on Reddit or Quora in your domain |
| Weeks 9–12 | Refresh any cornerstone content older than 12 months |
| Weeks 9–12 | Publish expert-led content with named contributors and sourced claims |
| Weeks 9–12 | Build 3–5 quality backlinks via PR or editorial placement |
| Ongoing | Review and refresh top pages every 6–12 months |
Final thoughts
LLM optimisation isn’t a replacement for SEO. The foundations, quality content, topical authority, technical health, credible backlinks, still matter and still carry over. What’s changed is the layer on top: how that content is structured, how specifically it speaks, and how consistently your brand appears across the broader web.
The brands that AI assistants learn to trust will increasingly become the brands that users discover first, without those users ever clicking through a search results page. Getting there is less about hacking an algorithm and more about being genuinely useful, specific, and present in the right places.
If you want to talk through where your brand stands on AI visibility and what’s worth prioritising, get in touch. It’s one of the more interesting strategic conversations we’re having with clients right now.
