LLM SEO Agency for Getting Cited as the Source

There's no evidence it does yet. No major AI engine has said it uses llms.txt to choose what to cite, and Google has said its search doesn't use it. It's cheap to write and harmless, so if you want one we'll add an accurate version. We just won't count it as progress, and we'd never let it stand in for crawler access, readable pages or outside mentions.

OpenAI documents GPTBot for training and OAI-SearchBot for search. Anthropic documents ClaudeBot for training and a separate crawler for search. PerplexityBot is documented as the crawler for that engine's search index. Blocking a search crawler can keep you out of that engine's answers, while blocking a training crawler mainly affects what future models learn. We set out the trade-off for each crawler and you decide.

Not directly, and nobody can. What a model absorbed before release only changes when a newer model is trained. You can shape what the next one sees by making the web describe you consistently across your own pages, review sites, roundups and community threads. Retrieval fixes change live answers much sooner, so we start there.

Retrieval fixes show first. Once a crawler can reach and read a page, engines that search at answer time can cite it after their next visit. Training-data change follows model release cycles, which can mean many months and isn't in anyone's control. We set the baseline in the first 30 days, report monthly, and give the first full read at day 90.

Same thing, different label. LLMO stands for LLM optimization, and vendors use it, LLM SEO and a few other names for making a model more likely to retrieve, trust and cite you. The work underneath doesn't change with the name: crawler access, readable pages, citable passages and outside corroboration, measured engine by engine.

HeyOz launched with a single homepage and now has 91 AI citations, from none, and is cited by 6 of 7 major AI answer engines. Viggle AI went from 4 to 279 pages cited in AI answers. Both engagements combined technical, content and authority work, measured against a fixed baseline set before anything shipped.

An answer can name your product from training data, or from a third-party page, without citing any page of yours. That's a mention, not a citation. Mentions show the model knows you. Citations send visits and let you check what was quoted. We track both separately and work on making your own pages the cited source for the prompts closest to a signup.

We rerun the same prompts every month on ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews, Google AI Mode and Copilot, and check Google and Bing rankings, since retrieval starts from them. For each prompt we log whether you're cited as a source, mentioned without a link or absent, along with the page or outside source the answer used.