How do I check if my SaaS shows up in AI search?
The fastest way is a five-prompt audit run across the four major AI tools. Ask each tool the same five questions a real buyer would ask, score the result, and you have a baseline you can improve against.
Don't outsource this to a tool the first time. Doing it manually for an hour teaches you more about your AI visibility than any dashboard.
Which AI tools should I check?
Four cover almost all B2B software research in 2026:
- ChatGPT with browsing enabled - by far the most-used.
- Perplexity - shows citations on every answer, so it tells you exactly which sources are being used.
- Google AI Overviews - increasingly the first thing buyers see on Google.
- Claude - growing fast in technical B2B audiences.
Gemini and Grok are worth checking if you serve their core audiences, but the four above will give you a representative picture.
What questions should I ask?
Use the same five prompts in every tool. They cover the spectrum of how buyers actually search.
Always start a fresh chat for each prompt. AI tools personalise based on conversation history, and you want clean, comparable results.
How do I score the results?
Use a simple 0-4 rubric. It's coarse enough to apply consistently and granular enough to track change over time.
For each prompt in each tool, record the score, what was said about you, and which sources were cited (Perplexity makes this obvious; for ChatGPT and Claude you'll need to ask follow-up questions like "where did you get that?").
How do I track this over time?
A spreadsheet is enough to start. Columns: date, tool, prompt, score, mentions, sources cited. Re-run the same five prompts every month. Aggregate your average score per tool and per prompt type. The trend matters more than any single data point.
Once your tracking is consistent and you understand what "moving" looks like, then consider a paid tool. Profound, Otterly and AthenaHQ all do AI visibility tracking. They are useful, but they don't replace the qualitative signal you get from running prompts yourself.
What does a "good" score look like?
For an established player in their category, an average of 2.5 or higher across all 20 cells (5 prompts x 4 tools) is healthy. A new or niche product should aim for 1.5 within six months of starting GEO work, and 2.5 within twelve.
Branded prompts (prompt 04 in the framework) should be a 3 or 4 in every tool within weeks of tightening your on-site content and Organization schema. If they aren't, your entity isn't being recognised yet - fix that first.
What should I do with the results?
Three things:
- Identify the pattern. Are you weakest on category prompts? Branded? Comparison? Each one points to a different fix - category weakness is usually an off-site authority problem; branded weakness is usually an on-site or schema problem.
- Capture the source list. For every cell scoring 2 or above, note the sources cited. That list IS your GEO target list - the third-party properties you need to be on.
- Pick one improvement per pillar. One technical fix, one on-site rewrite, one off-site outreach. Ship them this month. Re-run the audit. Compare.
Repeat monthly. The compounding effect is the whole game.