Everything we've learned about getting your brand named when buyers ask ChatGPT, Claude, Perplexity and Gemini. What actually works, what's a waste of time, and the exact week-by-week sequence to ship it — written by people who do this for a living.
robots.txt, llms.txt and Schema markupOr just scroll — the full text is below. We don't believe in paywalls for thinking.
Before any tactics, you need to understand how an LLM picks the brands it names. It is not the same as Google's ranking. It is not personalization. It is not paid placement. Here's what's actually happening.
When a buyer types "what's the best [thing] for [use case]?" into ChatGPT, the model produces an answer by combining three signals. If you don't show up in any of them, you don't exist.
The model was trained on a snapshot of the public internet up to its cutoff date. Whatever the internet "said" about your category — Reddit threads, Wikipedia, news, blogs, forums — is now baked into the model's instincts.
You cannot edit this. You can only influence the next training cycle by being talked about now.
Modern AI search (ChatGPT with browsing, Perplexity, Claude with web, Gemini) fetches the top 3–15 web pages at query time and synthesizes them into the answer. The brands that appear in those pages tend to make it into the response.
This is the lever you can pull this month. Get cited in the pages models fetch.
Inside any set of retrieved pages, models weigh them by trust. Wikipedia, Reddit (with caveats), major news outlets, established review hubs (G2, Capterra), and official sites carry more weight than random blogs or AI-generated junk.
Where you appear matters more than how many times you appear.
Everything in this cheatsheet is, in one way or another, a tactic for making one of those three things more true.
Ranked by real-world impact across the engagements we've run. Impact is "did doing this move share-of-AI-voice." Effort is "how long to ship the first instance."
| Tactic | Why it works | Effort | Impact |
|---|---|---|---|
| Get cited in “Best X for Y” listicles | LLMs synthesize 3–7 listicles per category query. If your name appears in 5 of them, it almost certainly appears in the answer. | Medium | Very high |
| Authentic Reddit presence in your category subs | Reddit is in every major model's training set and heavily retrieved at query time. A few genuine, helpful, non-promotional threads outrank a thousand backlinks. | Medium | Very high |
| G2 / Capterra / TrustRadius profile + reviews (B2B) | These are the trusted review hubs models lean on for software questions. Get to 10+ real reviews and you become "extractable." | Medium | Very high |
| “You vs. Competitor” comparison pages on your own site | Models fetch these at decision-time and extract concrete differentiators. Build one per major competitor. | Low | Very high |
| Wikipedia article (if you genuinely qualify) | Highest authority signal across every model. But only attempt if you pass notability — otherwise you waste editor goodwill. | High | Very high |
| Co-citation engineering | Engineer mentions where your name sits beside category leaders ("X, Y, and Z are the standouts…"). LLMs surface entities that travel together. | High | Very high |
| FAQ schema markup with direct Q&A pairs | Gives engines extractable, citable answers. Particularly powerful for capturing long-tail buyer questions. | Low | High |
| Press in mid-tier industry publications | Perplexity and Gemini both lean heavily on trade press. One feature in the right outlet > 50 backlinks from junk blogs. | Medium | High |
| Podcast appearances with full transcripts | Transcripts are indexed and quoted. A good 45-minute episode = ~10 pages of citable, declarative content about you. | Medium | High |
| HackerNews / Indie Hackers / Lobsters discussion (dev tools) | For technical products these forums are disproportionately weighted in both training and retrieval. | Medium | High |
| Direct, declarative homepage copy | "WeRankAnything is an AI ranking agency for B2B startups." Easy to extract, easy to quote. Avoid hype-stack openers. | Low | High |
| Original research / surveys / data | If you publish a real stat that becomes the canonical source, you become unskippable for any prompt that touches that topic. | High | High |
| YouTube videos with rich transcripts | YouTube transcripts are fetched and cited, especially by Perplexity and Gemini. Caption everything. | Medium | Medium-high |
| Schema.org Organization + Product/Service markup | Tells models exactly what you are, who founded you, what you sell. Cheap, fast, helpful. | Low | Medium |
| Newsletter mentions (Substack, beehiiv, mid-list) | Increasingly indexed. A single mention by a respected operator in your space is worth pursuing. | Low-medium | Medium |
| Stack Exchange / GitHub presence (technical products) | Domain-specific authority signal. Answering well-asked questions builds a quote-able paper trail. | High | Medium (niche) |
| Conference talks with public recordings | Transcripts make their way into training corpora. Good for thought-leadership-driven categories. | High | Medium |
| Strategic LinkedIn / X / Bluesky posts by experts | Increasingly indexed in real-time retrieval. Hard to engineer but worth seeding. | Medium | Medium |
| Wikidata entry (if you can't get Wikipedia) | Lower bar than Wikipedia. Establishes you as an entity that exists. Especially useful for Gemini. | Medium | Medium |
| llms.txt at site root | An emerging standard for telling LLMs what to crawl and how to summarize you. Low cost; not yet a major signal. | Low | Low-medium (emerging) |
| robots.txt that explicitly allows AI bots | Without this you're invisible to GPTBot, ClaudeBot, PerplexityBot. Table stakes. | Trivial | Critical (table stakes) |
| Open-source projects on GitHub (where relevant) | Establishes deep technical trust. For dev-tools, a popular OSS repo is the single highest signal. | High | Medium-high (niche) |
Half of the "AI SEO" advice on the internet is wrong, recycled SEO from 2016, or speculative. Here's what we've watched fail.
LLMs don't reward density. They reward citability. A page that says "best CRM" 40 times is less useful than one that says it once, clearly, with a real reason.
Every site in your category has one. Models recognize the pattern and skip them. Build something with actual data, opinion, or specificity instead.
Old-school SEO trick. Doesn't show up in LLM retrieval signals at all. A waste of budget.
Some AI bots don't render JS. If your positioning copy only exists post-hydration, you're invisible to half the crawlers that matter.
Mods will remove it, the community will flag your brand as spam, and you'll poison the well for legitimate engagement later.
Reverted within hours by editors. Often results in your page being protected against future legitimate edits. Hire a notable third party to do it properly or wait.
Detected, banned, profile damaged. Recovery is brutal. Don't.
The kind of "Top 10 SaaS Tools" listicles that sell slots for $300. LLMs do not trust these and largely ignore them.
"World's leading," "best-in-class," "revolutionary." Models discount these phrases and trust the surrounding text less. Write like a careful journalist instead.
If all your useful content lives behind a login, crawlers can't see it. You need a public, indexable layer — even if it's a summary.
"Acme Inc.," "Acme Tools," "AcmeApp" — pick one canonical name and use it everywhere. Entity confusion hurts your share-of-voice in ways that are very hard to debug.
Modern training data weights authority and recency over how old your domain is. The .com you registered in 2008 doesn't help you here.
No measurable effect on LLM ranking. Stop renting bestcrmforhealthcare.com.
This is the single most common mistake. You can't be retrieved if you can't be crawled. Unless you're a paywalled publisher, allowing AI bots is almost always the right call.
Publishing false comparisons hoping LLMs cite them. Models cross-check now. When you're caught, your domain trust drops across every engine.
Filtered as noise by editors and ignored by LLMs. If you want to be cited, write something worth citing.
Do this once and you'll never have to do it again. These are the table stakes — none of the strategy in this guide matters if you skip them.
robots.txt — allow every major AI botLive at yourdomain.com/robots.txt. Copy this as a baseline and merge with your existing rules.
# Allow the AI bots that matter
User-agent: GPTBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: Claude-User
Allow: /
User-agent: Claude-Web
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Perplexity-User
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: Applebot-Extended
Allow: /
User-agent: CCBot
Allow: /
User-agent: Bingbot
Allow: /
User-agent: Amazonbot
Allow: /
Sitemap: https://yourdomain.com/sitemap.xml⚠ Note: Google-Extended covers Gemini/Bard specifically. The regular Googlebot entry is separate and you probably already allow it.
llms.txt — a brief, Markdown summary of your siteLive at yourdomain.com/llms.txt. This is an emerging standard (low impact today, rising). Keep it short, declarative, and factual.
# WeRankAnything
> The AI ranking agency that gets B2B startups cited in
> ChatGPT, Claude, Perplexity and Gemini answers.
## What we do
- AI visibility audits across 4 major engines
- Mention engineering in the sources LLMs read
- Site retraining for clean LLM extraction
- Always-on prompt monitoring
## Pricing
- Audit: $2,500 one-time
- Starter: $3,500 / month
- Growth: $8,500 / month
## Founded
San Francisco, 2025. Founder: Jaisal R.Drop these in the <head> of the relevant pages. Most CMSes have a plugin for this if you don't want to write JSON-LD by hand.
Organization on your home page — name, logo, founder, sameAs (your social profiles).Product or Service on each offering page — name, description, price, provider.FAQPage on any page with Q&A — the single highest-ROI piece of structured data for AI search.Optional but valuable: Article on blog posts, Person on founder/author bios, BreadcrumbList on category pages.
Make sure you have a sitemap.xml referenced in your robots.txt, and every page has a self-referencing <link rel="canonical">. Models use canonical URLs to dedupe when they retrieve your content.
The top of every important page should contain at least one direct, declarative sentence about what you are. Not a hero pun. A sentence a model can cite verbatim.
Good: "WeRankAnything is a small agency that helps B2B startups appear in AI-generated answers from ChatGPT, Claude, Perplexity and Gemini."
Bad: "Transform your future-ready brand into the AI-native engine of tomorrow."
Each engine has its own retrieval personality. These are patterns we've observed across many engagements. They drift over time — check yourself monthly.
Win it by: getting into Wikipedia (or Wikidata), seeding sober Reddit discussion, building a Wikipedia-ish "About" section on your own site.
Win it by: getting cited in 5+ listicles, publishing dated comparison pages, ranking on the keywords Perplexity's web search uses (still partly SEO-shaped).
Win it by: publishing real research, getting cited by primary sources (docs, papers, well-respected blogs), keeping your own copy careful and quotable.
Win it by: not abandoning SEO. Ranking on Google for buyer queries, getting a Wikidata entry, structured data, and ensuring Google-Extended is allowed.
A realistic sequence you can run yourself, in-house, this month. Each week takes 5–10 hours of focused work.
robots.txt, llms.txt, and Schema markup from Section 4.Ignore traffic-shaped vanity metrics. These are the six numbers we put on our clients' dashboards.
What: The % of your tracked prompts where your brand appears at least once.
Why: The single best summary metric. If this goes up, you're winning.
What: When you're named, are you 1st, 2nd, or last in the recommended list?
Why: Buyers often click only the first one or two names. Position matters.
What: % of prompts where you appear alongside the category leaders.
Why: Co-citation is how LLMs decide who "belongs" in a category. Track who you travel with.
What: Across how many distinct sources are you being cited?
Why: One Reddit thread = fragile. Twenty diverse sources = durable.
What: Sessions where the referrer is chatgpt.com, perplexity.ai, claude.ai, or gemini.google.com.
Why: Direct proof the work is producing visits. Imperfect but real.
What: Ask every new customer "How did you find us?" Track responses mentioning an AI assistant.
Why: The most honest signal you'll get. Buyers will tell you ChatGPT recommended you, if you ask.
The cheatsheet is free for a reason: most of the impact is in the patient execution, not the knowing. If you want a small, careful team doing this every week, we're here.