Every time someone asks ChatGPT "what's the best CRM for startups" or "top running shoes under $150", a handful of brands get recommended and the rest don't. The question this article answers: what decides which bucket you're in?
We tested 20+ brand-level and category-level queries across ChatGPT's free, Plus, and API tiers, compared them to the same queries on Claude, Gemini, and Perplexity, and reverse-engineered the signals that consistently predict whether a brand appears in the answer.
Here are the 12 factors, ranked by our observed impact. Each one has a concrete action you can take this week.
ChatGPT's browsing tool leans heavily on Schema.org markup when it parses your site. A page with Organization, Product, FAQPage, or LocalBusiness JSON-LD gets extracted cleanly; a page without it gets messy HTML text that the model has to infer from.
Action: Add JSON-LD to your homepage, product pages, and at least one FAQ page. Validate with Google's Rich Results Test. Minimum viable set:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Your Brand",
"url": "https://yourbrand.com",
"logo": "https://yourbrand.com/logo.png",
"sameAs": ["https://twitter.com/you", "https://linkedin.com/company/you"]
}
</script>
Many sites block AI crawlers in robots.txt by default — sometimes inherited from a WordPress plugin, sometimes copied from a years-old template. If GPTBot can't fetch your site, ChatGPT can't cite you. Full stop.
Action: Open your robots.txt. Look for lines like User-agent: GPTBot followed by Disallow: /. Remove them. Explicitly allow:
User-agent: GPTBot Allow: / User-agent: ClaudeBot Allow: / User-agent: PerplexityBot Allow: / User-agent: Google-Extended Allow: /
Brands mentioned on 4+ external platforms (Reddit, Wikipedia, industry directories, review sites, major publishers) get cited at roughly 2.8x the rate of single-platform brands. This is the AI-search equivalent of a backlink profile, but the signal is "are we a thing the internet consistently talks about" rather than raw link equity.
Action: Inventory where your brand appears. If you're only on your own site + a Google Business Profile, you have a presence problem. Target: 4+ platforms with non-thin mentions within 6 months. Good targets: Wikipedia, Reddit (organic, not spam), G2/Capterra for SaaS, Yelp for local, industry-specific directories.
ChatGPT loves direct Q&A format because it can extract and cite answers verbatim. Pages with FAQPage schema outperform unstructured content on equivalent topics — they become the literal source of the answer.
Action: Pick your top 10 customer questions (look at support tickets, sales calls, "People Also Ask" on Google). Build a /faq page. Each Q&A gets FAQPage schema. This alone can move SOV 3-5 points.
The emerging "readme for AI" standard. A simple markdown file at https://yoursite.com/llms.txt that tells AI crawlers: "here's who we are, here are our most important URLs, here's what we do in one paragraph." Not all engines respect it yet, but Perplexity and Claude's internal retrieval both index it, and adoption is growing.
Action: Create llms.txt. Keep it under 200 lines. Structure:
# Your Brand Name > One-sentence positioning line. ## About Short paragraph. Who you are, who you serve, what's different. ## Key Pages - [Homepage](https://yoursite.com) - [Product](https://yoursite.com/product) - [Pricing](https://yoursite.com/pricing) - [FAQ](https://yoursite.com/faq) ## Optional More context, case studies, team.
ChatGPT heavily weights review-based sources (Amazon, Yelp, G2, Trustpilot, Capterra). A brand with 500 recent reviews averaging 4.7 will almost always outrank one with 2,000 stale reviews averaging 4.2. Recency matters as much as volume.
Action: Audit your review footprint on the 2-3 platforms that matter for your category. Fix any that are below 4.5. Set up an automated review-request flow for customers at a positive moment (post-delivery, post-onboarding, post-renewal).
Thin content rarely gets cited. ChatGPT consistently picks comprehensive pages over listicle-style "top 10" posts for category queries. 2,000+ word pages that genuinely answer the question outrank 500-word SEO-farm posts, even when the latter ranks higher on Google.
Action: Identify the 5 highest-intent queries in your category. Write one deep article (2,500+ words) per query. Structure with H2/H3 hierarchy, include tables and lists (AI loves these for extraction), cite primary sources.
Wikipedia is a training-data goldmine for every LLM. A clean, notability-passing Wikipedia page is the closest thing to a cheat code for ChatGPT's base knowledge. For brands that qualify (notable products, companies with press coverage, public figures), this is worth the editorial effort.
Action: If your brand has press coverage from major outlets, work with a writer who knows Wikipedia's sourcing rules. Don't self-edit — get the page created organically via a neutral third party. Focus on sourceable facts, not marketing copy.
When ChatGPT browses your site, the first signals it reads are <title>, <meta description>, and og: tags. These should state what you are, what you sell, and who you serve in under 160 characters — because those 160 characters are often what the model summarizes from.
Action: Audit your homepage meta. Title should include brand + category + differentiator. Description should answer "what would a customer need to know in one sentence?" Not your tagline — your actual value prop.
ChatGPT extracts sentences and paragraphs verbatim. Pages that lead with a direct answer (then expand) outperform pages that bury the answer. If the question is "what's the best CRM for a 10-person startup," the article's first sentence should name a CRM and say why — not set up 3 paragraphs of context.
Action: For every pillar article, rewrite the first paragraph so the answer is in the first sentence. Treat the rest as the "why." Use the inverted-pyramid journalism pattern.
ChatGPT reportedly weights Reddit highly as a source for real-user opinion — especially for product comparison and recommendation queries. Organic mentions in relevant subreddits (where humans actually recommend you) drive citations.
Action: Don't spam. Do find 3-5 subreddits where your ICP already talks. Participate genuinely for 60+ days before mentioning your product. Build a pattern where organic recommendations happen naturally.
ChatGPT's browsing tool biases toward recently-updated content for time-sensitive queries. A 2024 "best tool" article gets outranked by a 2026 one even if the 2024 is more comprehensive. For category pages, an update timestamp in the next 12 months is table stakes.
Action: Pick your top 20 pages by organic traffic. Refresh each one quarterly with a visible "Updated [month] [year]" stamp. Add genuinely new information — don't just touch the date.
Three tactics we tested extensively that had near-zero impact on ChatGPT visibility:
If you want the fastest possible improvement in ChatGPT visibility, this is the sequence:
| Week | Action | Expected SOV lift |
|---|---|---|
| Week 1 | Baseline audit. Fix robots.txt to allow GPTBot/ClaudeBot/PerplexityBot. Add Organization + Product Schema JSON-LD. | +2-4 points |
| Week 2 | Ship llms.txt. Build FAQ page with FAQPage schema (10 questions minimum). Rewrite homepage meta tags. | +3-6 points |
| Week 3 | Publish one 2,500-word pillar article on your top category query. Update 5 existing articles with fresh "Updated [date]" stamps and new information. | +4-8 points |
| Week 4 | Review audit: fix your 2-3 most important review platforms (volume + recency). Claim/update directory listings (Yelp, G2, industry-specific). | +3-5 points |
| Total | 30 days of focused work | +12-23 SOV points |
These are typical gains. Brands starting from a very low baseline often see bigger jumps; brands already above 50 SOV see smaller incremental lifts because they've already captured the easy wins.
You can't improve what you don't measure. Track monthly:
A tool like Teehoo Martech's free Brand Scan will do steps 1-3 in ~2 minutes. For step 4, you'll want to manually ask ChatGPT your target queries and note the citations.
Free 2-minute scan across ChatGPT, Claude, Perplexity, and Gemini. No signup needed.
Scan Your Brand or Run a full GEO auditRelated reading
· GEO for Agencies: The Complete Playbook
· AI Share of Voice Tracking: A Complete Guide
· The AEO Audit Checklist: 14 Points
· ChatGPT vs Perplexity vs Gemini: Which AI Matters Most?
About Teehoo Martech
AI visibility platform for brands and agencies. Track how ChatGPT, Claude, Perplexity & Gemini recommend you — and the content gaps between you and your competitors. Learn more.