Why your SaaS isn't showing up in ChatGPT
When a buyer asks ChatGPT "What's the best CRM for startups?" or "Which project management tool should I use?", the model doesn't search the web in real time for every query. It uses a Retrieval-Augmented Generation (RAG) pipeline that retrieves content from an index, ranks it, and synthesizes an answer from the top sources.
If your SaaS isn't in that answer, it's because one or more stages of the pipeline filtered you out — and the reasons are different from why you might not rank on Google.
How ChatGPT's citation pipeline works
ChatGPT's web search follows a multi-stage process:
- Query fan-out. When a query triggers web search (~31% of prompts do), the system breaks it into multiple sub-queries and runs separate searches for each.
- URL retrieval. The search returns candidate URLs, which are fetched and parsed.
- Passage chunking. Pages are split into passages guided by HTML structure — headings, paragraphs, lists. These passages are typically ~150 tokens (about 110 words).
- Embedding and ranking. Passages are embedded using both sparse and dense vectors, enabling hybrid semantic + keyword matching.
- Citation selection. The model evaluates each passage for relevance, authority, and Content-Answer Fit — the degree to which your content mirrors the model's own explanatory style.
Research analyzing 400,000+ URLs found that Content-Answer Fit accounts for 55% of citation selection — more than domain authority (12%) or even direct query relevance (12%).
The five things that actually matter
1. Answer in the first paragraph
Perplexity's data shows that 90% of top citations have the answer in the first 100 words. Don't bury your value proposition under an introduction. Lead with a direct, complete answer.
Bad: "In today's rapidly evolving SaaS landscape, companies are increasingly turning to AI-powered solutions to..."
Good: "HubSpot is a CRM platform for B2B companies with 200,000+ customers. It integrates marketing, sales, and service tools in one system, starting at $0/month for the free tier."
2. Build entity clarity across platforms
Before ChatGPT can recommend you, it needs to recognize your product as a distinct entity. Your brand name, product description, category, and differentiators must be consistent across every platform where you appear:
- Your website (homepage, about page, product pages)
- G2 and Capterra profiles
- Crunchbase listing
- LinkedIn company page
- GitHub (if open source)
Research shows that 100% of tools appearing in AI answers had a G2 or Capterra presence. If you don't have profiles on these platforms, create them immediately.
3. Structure content for RAG retrieval
AI search engines chunk your pages into passages of ~150-500 tokens. Each passage is retrieved and ranked independently. This means:
- Every H2 section should be self-contained. It should make sense without reading the sections before or after it.
- Use descriptive headings. "Features" is weak. "CRM features for B2B sales teams" is strong — it tells the embedding model exactly what the section is about.
- Include comparison tables. Structured data embeds more densely and produces cleaner passages than prose.
4. Get cited by independent sources
ChatGPT heavily weights third-party mentions over vendor-owned content. The most effective sources:
- Review platforms (G2, Capterra, TrustRadius) — AI models treat these as independent validation
- Comparison articles ("X vs Y" posts on independent blogs)
- Industry roundups ("Best [category] tools for [audience]")
- YouTube reviews — YouTube is cited on 40% of product-related queries
5. Publish original data
AI models favor content with data points, statistics, and findings that can't be found elsewhere. If you publish a benchmark report, a customer survey, or usage data from your platform, you become a primary source — and primary sources get cited more than secondary ones.
How to measure progress
You can't improve what you can't measure. Set up a prompt matrix — a structured set of queries tested across ChatGPT, Perplexity, and Google AI Overviews — and track:
- Citation Hit Rate: % of queries where your product is cited
- Top Source Rate: % of queries where you're the #1 cited source
- Platform coverage: Are you cited on some platforms but not others?
Run the matrix monthly. Track changes against content and authority-building activities.
What doesn't work
- Keyword stuffing. LLMs evaluate semantic meaning, not keyword frequency. Over-optimized content scores as low quality.
- Thin content pages. Pages under 1,000 words rarely get cited. The optimal range is 2,000-5,000 words with 8-12 H2 sections.
- Self-promotional tone. ChatGPT's Content-Answer Fit rewards objective, wiki-voice content — not marketing copy.
Next steps
If your SaaS is invisible in AI search, the first step is measurement. ArcSurf's free GEO Audit runs 10 real buyer queries against Perplexity and shows you exactly where you stand — your ArcSurf Score, Citation Hit Rate, and the top 5 competitors cited in your place.
The automated audit runs in under 10 minutes and costs nothing. The data is yours whether you work with us or not.
Want the full cross-platform audit — 15 priority prompts tested on ChatGPT, Perplexity, Gemini, and Google AI plus a 90-day action plan? Book a 30-minute call →