Make AI describe
your brand
correctly.
LLM Optimization shapes how large language models — the AI behind ChatGPT, Claude, Gemini and Perplexity — understand, describe and recommend your brand. Reinforce Lab makes your brand facts consistent everywhere models learn from and look things up, then tests what the models actually say.
How does an AI model know about your brand?
There are two routes, and they need different fixes. The AI companies document both.
What it learned in training
Models are trained on data up to a cut-off date. OpenAI's own help pages say its models "do not incorporate information about events beyond that, unless tools are used." Whatever the web said about you before that date is baked in.
FixConsistent, accurate facts about you across the web — so the next training round learns the right story.
What it looks up live
With search switched on, tools such as ChatGPT search, Perplexity and Gemini with Google Search grounding fetch current pages and cite them. Google says grounding lets Gemini cite "verifiable sources beyond its knowledge cutoff."
FixClear, current, crawlable pages that state your facts plainly — and a crawler policy that lets search bots in.
Which AI crawlers should you allow?
Not all AI crawlers do the same job. Some collect training data, some power AI search, and some fetch a page only when a person asks. Blocking the wrong one can remove you from AI answers.
| Company | User agent | Used for | What the company says |
|---|---|---|---|
| OpenAI | GPTBot | Training | Collects content that may be used to train OpenAI’s models. Source |
| OpenAI | OAI-SearchBot | AI search | Surfaces sites in ChatGPT search. Sites that opt out are not shown in ChatGPT search answers. Source |
| OpenAI | ChatGPT-User | User request | Visits a page when a user’s question needs it; not an automatic web crawler. Source |
| Anthropic | ClaudeBot | Training | Collects web content that could contribute to training Claude models. Source |
| Anthropic | Claude-SearchBot | AI search | Crawls to improve the quality of Claude’s search results. Source |
| Anthropic | Claude-User | User request | Fetches pages when a user asks Claude a question. Source |
| Perplexity | PerplexityBot | AI search | Surfaces and links sites in Perplexity results; not used to train foundation models. Source |
| Perplexity | Perplexity-User | User request | Visits a page to answer a user’s question; not used for training. Source |
| Google-Extended | Training & grounding | A control token for use in Gemini training and grounding. It does not affect inclusion or ranking in Google Search. Source |
Summarised from each company's own crawler documentation, checked 29 September 2026. Crawler names and rules change — we re-check them for every client. Google's AI Overviews and AI Mode use pages indexed for Google Search; Google-Extended does not control that.
Why do AI models get brands wrong?
Models repeat what the web tells them. When the web is unclear, the answer is too.
Different stories in different places
Your site, profiles and directories describe you in different words — or list different services.
Old information still online
Past services, old prices or former addresses keep being repeated long after they changed.
Someone else shares your name
Similar brand names get merged, and their facts end up in your description.
Nothing clearly states the facts
Without a clear About page and structured data, models have to guess who you are and what you do.
Few independent mentions
When trusted third parties rarely mention you, models know little — or lean on competitors.
The wrong crawlers blocked
A blanket block on "AI bots" can shut out the search crawlers that would have cited your current facts.
How does LLM Optimization work?
Five steps, starting with what the models say about you today.
Entity baseline
Ask each model the questions buyers ask — who you are, what you do, how you compare — and record every answer.
Trace the errors
For each wrong or missing fact, find the pages and profiles it most likely comes from.
Set the source of truth
One clear description, a facts-rich About page, and Organization schema that links your official profiles.
Align the web
Update profiles, directories and listings to match, correct outdated pages, and agree your AI-crawler policy.
Re-test & monitor
Ask the same questions again on a schedule, and catch new errors before buyers do.
What you get.
- Model description report — how each AI model describes you today, answer by answer.
- Error log — every wrong, outdated or missing fact, with its likely source.
- Brand fact sheet — the agreed name, description, services and key facts, in one place.
- About page & schema — a facts-first About page and Organization markup linking your official profiles.
- Profile alignment — the profiles and listings to update, with the exact wording.
- AI-crawler policy — which crawlers to allow or block, set in your robots rules.
- llms.txt — an optional map of your key pages for AI agents.
- Re-test reports — what changed in each model's answers since the baseline.
How do we measure LLM Optimization?
By what the models say, on a fixed set of questions, over time.
Description accuracy
Share of answers that describe you correctly.
Consistency across models
Whether different AI tools tell the same, correct story.
Entity recognition
Whether models know you exist, and put you in the right category.
Recommendation presence
How often you are suggested for the needs you serve.
Errors resolved
Wrong facts traced, corrected at the source and no longer repeated.
Nobody can edit an AI model for you.
No agency can change what a model learned or make it say something. What we can do is fix what models read: your own pages, your profiles and the sources that describe you — and keep the crawlers that fetch fresh facts able to reach you. AI tools that search the web pick those fixes up first; trained memory follows as models are retrained.
Who is LLM Optimization for?
Brands that AI tools describe wrongly, vaguely or not at all — and teams in regulated fields, where a wrong description is a real risk.
About LLM Optimization.
What is LLM Optimization?
LLM Optimization is the work of shaping how large language models — the AI behind tools such as ChatGPT, Claude, Gemini and Perplexity — understand, describe and recommend a brand. It makes the facts about a brand consistent and easy to find everywhere models learn from and look things up, then checks what the models actually say.
How is LLM Optimization different from GEO and AI Search Optimization?
AI Search Optimization is the full program for being visible in AI search. GEO works on pages and passages so they get cited in answers. LLM Optimization works on the brand itself as an entity: who you are, what you do and who you serve, stated the same way everywhere — so models describe you correctly even when no page is cited.
Can you fix wrong information ChatGPT or another AI says about us?
Not by editing the model — no one outside the AI company can do that. We find where the wrong information comes from, correct it at the source, publish clear and current facts on your own site, and re-test. Tools that search the web can pick up corrections quickly; what a model learned in training changes only when it is retrained.
Should we block AI crawlers?
It depends on the crawler. Some collect data for training models, others fetch pages for AI search answers. OpenAI, for example, says sites that opt out of its search crawler will not be shown in ChatGPT search answers, while its separate training crawler can be blocked without that effect. We help you set a policy per crawler, based on what you want.
What is llms.txt, and do we need one?
llms.txt is a proposal, published by Jeremy Howard in September 2024, for a plain-text file at a site’s root that points AI agents to a site’s most useful pages. It is not an official standard and no major AI company has said it relies on it, so we treat it as a low-cost extra, not a ranking factor.
How long does LLM Optimization take?
Corrections on your own site and profiles can be picked up quickly by AI tools that search the web. Changes to what a model learned during training take longer, because they depend on the next training cycle. We baseline how each model describes you first, then report changes against that baseline.
What does AI say about you today?
The free Search Authority Diagnostic shows how AI tools describe your brand, where the facts go wrong, and what to fix first.