Key takeaways
- When someone asks ChatGPT to recommend software, the model secretly fires 8–10 parallel sub-queries ("best X reviews 2026", "X vs Y", "X pricing", "X complaints") before generating its answer. This is called query fan-out.
- 95% of fan-out sub-queries show zero monthly search volume in traditional keyword tools, which means most SaaS companies have no idea these searches are happening.
- A DerivateX study found that 44% of B2B SaaS companies are invisible in AI-assisted buyer searches -- not because their product is weak, but because their content doesn't match the sub-queries AI models run.
- Winning in AI search requires covering the full fan-out tree: reviews, comparisons, pricing, limitations, use cases, and freshness signals -- not just your homepage.
- Dedicated AI visibility platforms can map your fan-out coverage and show exactly where competitors are visible and you're not.
What query fan-out actually is
Most SaaS marketers think about AI search the way they think about Google: someone types a query, results appear, maybe your site shows up. That mental model is wrong, and it's costing companies real pipeline.
When a buyer types something like "what's the best CRM for a 50-person sales team" into ChatGPT, the model doesn't just retrieve one answer. It fans out. Before generating a response, it runs somewhere between 8 and 10 parallel sub-queries -- each one probing a different angle of the question. Research analyzing 72,000+ AI-generated queries across 8,700+ prompts found this pattern is consistent across ChatGPT and Gemini.
Those sub-queries look something like this:
- "best CRM for small sales teams 2026"
- "HubSpot vs Salesforce vs Pipedrive comparison"
- "CRM reviews Reddit 2026"
- "Pipedrive pricing plans"
- "HubSpot complaints limitations"
- "best CRM free trial"
- "CRM for 50 person team pros and cons"
The buyer never sees these queries. They just see the final answer. But every one of those sub-queries is a gate your brand has to pass through to appear in that answer.
The implication for SaaS companies is uncomfortable: you can have a great product, a well-optimized website, and strong SEO -- and still be invisible in AI search because you haven't addressed the specific sub-queries the model runs during evaluation.
Why software evaluation prompts fan out more than other queries
Not all prompts trigger the same level of fan-out. Software buying decisions are high-stakes, which means AI models apply more due diligence before answering. The model is essentially mimicking what a careful buyer would do: check reviews, compare options, look at pricing, find complaints.
Fan-out frequency and depth correlates with:
- Purchase complexity: Enterprise software prompts fan out more than "what's a good note-taking app" prompts
- Category competition: Crowded categories (CRM, project management, marketing automation) trigger more comparison sub-queries
- Price point: Higher-priced tools trigger more "pricing", "ROI", and "worth it" sub-queries
- Recency sensitivity: Software categories change fast, so models inject "2025" or "2026" into sub-queries to find fresh signals
A study of 5M+ ChatGPT fan-out queries found that the terms "best", "reviews", and "2026" are systematically injected into sub-queries for software evaluation prompts. The model isn't waiting for users to add those qualifiers -- it adds them automatically.
The 44% visibility gap in B2B SaaS
A DerivateX study found that 44% of B2B SaaS companies are effectively invisible when buyers use AI to research software. That's not a small rounding error -- nearly half of the market has no presence in the channel that's increasingly driving purchase decisions.
The reason isn't that these companies have bad products. It's that their content doesn't match the sub-queries AI models run. They have homepage copy, feature pages, and maybe a blog -- but they're missing:
- Comparison pages that address "X vs Y" queries
- Review aggregation signals that satisfy "reviews" sub-queries
- Transparent pricing content that addresses "cost" and "pricing" sub-queries
- Honest limitation or "cons" content that satisfies "complaints" sub-queries
- Recent content with freshness signals that satisfy "2026" sub-queries
The model can't cite what it can't find. And if it can't find your content across multiple sub-query angles, it won't include you in the final answer -- even if you're the best option.
How fan-out shapes the typical SaaS evaluation prompt
Let's walk through a concrete example. A buyer prompts: "I need project management software for a remote engineering team of 20 people."
Here's roughly what the fan-out tree looks like:
| Sub-query type | Example sub-query | What it's checking |
|---|---|---|
| Category + year | "best project management software for engineering teams 2026" | Freshness, category fit |
| Comparison | "Linear vs Jira vs Asana for engineering teams" | Competitive positioning |
| Reviews | "Linear reviews Reddit 2026" | Social proof, real user sentiment |
| Pricing | "Linear pricing plans per user" | Budget fit |
| Limitations | "Linear cons limitations complaints" | Risk signals |
| Use case fit | "project management software remote engineering team" | Specific fit |
| Alternatives | "Jira alternatives for small engineering teams" | Competitive alternatives |
| Free trial | "Linear free trial" | Conversion signals |
Each row is a separate query the model runs. Your brand needs to appear in enough of them to make the final answer. If you're only visible in one or two, you're likely not getting cited.
What this means for your content strategy
The fan-out model completely changes how SaaS companies should think about content. Traditional SEO optimizes for high-volume head terms. Fan-out optimization requires covering the full evaluation tree -- including queries that show zero search volume in any keyword tool.
Cover the comparison layer
"X vs Y" content is one of the most reliably triggered fan-out sub-queries for software. Every major competitor pairing deserves its own page. Not a thin "we're better because X" page -- a genuinely useful comparison that acknowledges where each tool is stronger.
AI models are good at detecting promotional content. A comparison page that only says positive things about your product and negative things about competitors reads as biased and gets deprioritized. Pages that acknowledge real tradeoffs get cited more often.
Address limitations honestly
This one feels counterintuitive but it's important. The model runs "complaints" and "limitations" sub-queries specifically to balance its recommendation. If the only content about your product's limitations lives on competitor sites and Reddit threads, that's what gets cited.
Publishing an honest "limitations of [your product]" or "when [your product] isn't the right fit" page gives the model something authoritative to cite when it runs those sub-queries. It also builds trust with buyers who find it directly.
Build freshness signals into your content
The systematic injection of "2026" into fan-out sub-queries means dated content gets filtered out. A comparison page from 2023 that hasn't been updated is a liability. Build a content calendar that treats freshness as a ranking signal -- update key pages quarterly, add "updated [month] [year]" signals, and publish new content that references current product capabilities.
Target the "free trial" and "pricing" sub-queries
These are conversion-intent sub-queries that the model runs to help buyers understand the path to purchase. Your pricing page and trial signup flow need to be crawlable and clearly structured. If the model can't find clear pricing information, it may cite a competitor that has it.
How to track your fan-out coverage
This is where most SaaS marketing teams hit a wall. Traditional SEO tools don't show you fan-out sub-queries because they have zero search volume. You can't track what you can't see.
There are a few approaches:
Manual prompt testing
You can manually run evaluation prompts in ChatGPT, Perplexity, and Gemini and observe which brands get cited. This gives you qualitative signal but doesn't scale. You can't manually test hundreds of prompts across multiple AI models and track changes over time.
AI visibility platforms
Dedicated platforms have emerged specifically to track how brands appear in AI search responses. They run prompts at scale, record citations, and show you where competitors are visible and you're not.
Promptwatch goes further than most -- its Answer Gap Analysis shows exactly which prompts competitors are being cited for that you're missing, then its Content Agents help you create content to fill those gaps. The cycle from gap identification to content creation to citation tracking is built into the platform, which is what separates it from monitoring-only tools.
For teams that want to start simpler, a few other tools are worth knowing about:
Here's how the main approaches compare:
| Approach | Scale | Fan-out visibility | Gap analysis | Content creation | Cost |
|---|---|---|---|---|---|
| Manual prompt testing | Low | Partial | No | No | Free |
| Promptwatch | High | Full | Yes | Yes (Content Agents) | From $99/mo |
| Profound | High | Partial | Limited | No | Higher price point |
| AthenaHQ | Medium | Partial | Monitoring only | No | Mid-range |
| Rankscale | Medium | Basic | No | No | Lower |
The fan-out content audit: where to start
If you're starting from scratch, here's a practical sequence:
Step 1: Map your evaluation prompts. Write down every way a buyer might ask an AI to find software like yours. Include category-level prompts ("best CRM for startups"), use-case prompts ("CRM that integrates with Slack"), and comparison prompts ("HubSpot alternatives").
Step 2: Run those prompts across ChatGPT, Perplexity, and Gemini. Record which brands appear in each response. Note which brands appear consistently across multiple prompts -- those are your visibility benchmarks.
Step 3: Identify your sub-query gaps. For each prompt where you don't appear, think through the fan-out tree. Which sub-query types are you missing content for? Reviews? Pricing? Comparisons? Limitations?
Step 4: Prioritize by prompt volume and competition. Not all evaluation prompts are equal. Focus first on the prompts with the highest buyer intent and the most realistic path to visibility.
Step 5: Create content that addresses specific sub-queries. Each piece of content should clearly address one or more fan-out sub-query types. A comparison page should be structured so the model can extract a clear answer to "X vs Y". A pricing page should have clear, crawlable pricing information.
Step 6: Track changes over time. This is where manual testing breaks down. Use a platform that monitors your citation rates across prompts and models so you can see whether your content changes are working.
What "winning" a fan-out looks like in practice
There's a documented case of a B2B SaaS company going from 575 to 3,500+ trials per month after implementing an answer engine optimization strategy focused on AI search visibility. The core of the approach was covering the full evaluation prompt tree -- not just optimizing for traditional search terms.
The pattern that works:
- Appear in category-level prompts ("best [category] software")
- Appear in comparison prompts ("X vs [your product]")
- Have review signals that the model can find and cite
- Have clear pricing information
- Have content that addresses limitations and use-case fit
When you cover enough of the fan-out tree, the model has enough confidence to include you in its answer. When you're missing large sections of it, the model defaults to competitors who have covered those angles.
The Reddit and YouTube factor
One thing that surprises SaaS marketers: AI models don't just cite company websites. They actively look for third-party signals -- Reddit threads, YouTube reviews, forum discussions -- when running review and sentiment sub-queries.
A Reddit thread titled "honest review of [your product] after 6 months" can be more influential in AI search than your own review page. The model treats it as an independent signal, which is exactly what it's looking for when it runs "reviews" sub-queries.
This means your fan-out strategy needs an offsite component. Monitor what's being said about your product on Reddit and YouTube. Engage authentically in relevant discussions. Make sure the third-party content about your product is accurate and reasonably positive -- because the model is reading it.
Platforms like Promptwatch track offsite citations specifically, showing which Reddit threads and YouTube videos are influencing your AI visibility. That's data you can't get from traditional SEO tools.
Building a sustainable fan-out strategy
Fan-out tracking isn't a one-time audit. The sub-queries AI models run evolve as the models update, as competitors publish new content, and as buyer behavior shifts. What works today may need adjustment in three months.
The SaaS companies that will win in AI search are the ones that treat it as an ongoing program, not a project. That means:
- Regular prompt monitoring across multiple AI models
- Quarterly content freshness updates on key pages
- Continuous competitive monitoring to catch when competitors gain visibility
- A content pipeline that can respond to newly identified gaps
The infrastructure for this didn't exist two years ago. Now it does -- and the gap between companies that use it and companies that don't is already showing up in trial volumes and pipeline data.
Query fan-out is how AI models do their due diligence. The SaaS companies that understand this and build content to match the full evaluation tree are the ones that will show up in the answers buyers are already getting.