AEO for SaaS companies
Software buyers ask AI assistants to shortlist tools before they visit a single vendor site. This page explains why AI visibility behaves differently for SaaS, what is worth measuring, and the parts you can check without any tool at all.
Why AI visibility behaves differently here
Buyers ask for shortlists, and shortlists are assembled from third-party pages. Review sites, comparison posts and documentation feed the answer alongside your homepage, so your own site is one voice among several you do not control.
The 'alternatives to X' question routes real evaluations, and whether you appear in that answer differs engine by engine. Without sampling actual answers per engine, you cannot know where you stand on any of them.
Your facts expire quickly. Pricing, plan limits and feature availability change often, while answers lean on whatever was crawled or trained earlier, so assistants can keep repeating numbers you retired months ago.
Category language is contested. An assistant may file your product under a different category name than the one your positioning uses, and that mismatch stays invisible until you read real answers.
What is worth measuring
Each of these is a real measurement, not a modeled score. How each one is produced, sample sizes and intervals included, is documented on how we measure.
Citation share on shortlist prompts
Sample the shortlist and category questions your buyers actually ask across Perplexity, Google AI Overviews and ChatGPT. Every reported share carries its sample count and a 95% Wilson interval, so per-engine differences are measured rather than guessed.
The gap list
Shortlist and alternatives prompts where rivals are cited and you are not, taken from the same sampled answers. This is the observed version of the question every positioning meeting argues about.
Answer accuracy against your own facts
Declare your pricing, plan limits and feature facts once, and sampled answers are checked against them. Contradictions are flagged, and a claim you have since changed is marked as expired rather than silently treated as wrong forever.
Crawler access on marketing site and docs
Read robots.txt per named AI crawler on the surfaces that matter. Docs often live on their own subdomain with their own robots.txt, and that file gates what assistants can learn about your product in depth.
AI referral traffic
Sessions that arrive from AI assistants, tracked as their own referral class, so you can see whether being cited actually sends people, instead of assuming it does.
What you can do yourself, starting today
- 1
Run your own evaluation prompts across ChatGPT, Perplexity and Google, repeatedly, and write down who appears. Per-engine differences are normal; a single run on a single engine tells you almost nothing.
- 2
Keep one canonical, crawlable pricing and limits page, and retire stale numbers everywhere else. Assistants repeat old figures longest when the current ones are hard to find.
- 3
Publish the comparison your buyers already ask assistants for, and be honest about what rivals do better. An evasive comparison page is easy for both readers and engines to discount.
- 4
Check robots.txt on your marketing site and on your docs subdomain separately. They are different files, and it is common for one to have been configured years ago and never revisited.
- 5
When you want the shortlist question answered with sample sizes instead of anecdotes, run our free measured audit at /audit. No signup required.
Every number the audit returns is measured or marked not measured. Before you trust any tool with this job, ours included, read measured vs asserted.
AEO guides for other verticals: Shopify and ecommerce · Local businesses