Meet your new crawlers
Your logs already show them: GPTBot (OpenAI), Google-Extended and Gemini's fetchers, PerplexityBot, ClaudeBot (Anthropic). They do two jobs — collecting training data for future models and fetching pages live to ground current answers. The second job is the one you feel week to week: a page fetched today can be cited tomorrow.
Two habits distinguish them from Googlebot. They're less patient with client-side rendering — content that only exists after JavaScript executes is content some fetchers never see. And they read fewer pages per visit, which raises the stakes on each page being self-explanatory rather than depending on six clicks of context.
What actually moves answers
Pages that answer questions directly
Retrieval rewards extraction-friendly writing: the answer in the first two sentences, one topic per page, headings that match how customers phrase questions. The "short answer" pattern at the top of this post isn't house style for fun — it's the shape engines quote. If your key facts live in a PDF, a webinar, or a carousel, they functionally don't exist.
Structured data
Schema.org markup — Organization, Product, FAQPage, Article — gives engines machine-readable facts to anchor on: who you are, what you cost, what category you're in. It's the same markup search rewarded for a decade, now with a second customer. Among technical GEO tasks, this one has the clearest payoff for entity clarity, especially if your brand name is ambiguous.
Facts that agree with each other
Crawlers read your whole footprint. Pricing that differs between your homepage and your directory listings isn't just untidy — it's a coin-flip you've handed the engine. Same name, same description, same numbers, everywhere. Boring, effective, chronically underdone.
Where llms.txt honestly stands
llms.txt is a proposed convention: a markdown file at your site root that hands language models a curated map of your important pages. The case for it: costs an hour, can't hurt, and some fetchers do request it. The case for calibration: no engine has committed to honoring it, and it will not rescue content that's unreadable once fetched. Ship one — Brandflare does — but if it's above structured data on your to-do list, the list is upside down.
The 80/20 of technical GEO
- Server-render the content that matters; don't make fetchers execute JavaScript to find your pricing.
- Put the answer at the top of the page, not after the brand story.