AEO for SaaS: Find the Pages AI Cites Before You Fix Yours
AEO for SaaS, answer engine optimization for software-as-a-service firms, starts by finding which pages AI engines cite for your category, since review sites and community pages are among the most cited sources.
By Harjot Chopra, Co-Founder, PageLens.ai · Updated 18 Nov 2026

Key takeaways
- AEO for SaaS starts with a source-gap audit: log which pages ChatGPT, Perplexity and Google AI Mode cite for your category prompts, and which of them you are missing from.
- Review sites and third-party lists are among the most-cited sources, so check for a gap there before rewriting your own pages (our inference).
- In our November 17, 2026 fetch of 60 SaaS sites, few robots.txt files blocked AI crawlers (robots.txt only; edge blocks are not visible to it). 13 of 43 pricing pages showed fewer than three dollar figures in plain HTML, though some may be sales-led by design.
- Run three passes of the sheet by hand (135 answers), then track the same prompts on a schedule so you can tell a real change from noise.
In this article
- 1What does AEO for SaaS involve, and where should you start?
- 2Which pages you don't own shape a software shortlist?
- 3Is your own site blocking AI crawlers or hiding its prices?
- 4How do you run a source-gap audit for your SaaS category?
- 5Which 15 prompts should you run?
- 6What do you log for each answer?
- 7Which three numbers do you calculate?
- 8What should you fix first once the log shows your gaps?
- 9Which of your own SaaS pages should you fix next?
- 10How do you tie AI answers to trials?
- 11How do we find and close source gaps at PageLens.ai?
- 12Frequently asked questions
Software buyers already ask chatbots for shortlists. In G2's Buyer Behavior Report, published July 22, 2026, 82% of more than 1,000 B2B software buyers had sourced software recommendations from an AI chatbot in the previous 24 months.
The instinct is to rework your own site: answer-first paragraphs, FAQ markup, a comparison table. That work matters, but it acts on only one of the sources an answer draws on.
The first job is to measure which pages the engines cite. Below is a 15-prompt source-gap sheet, with three passes you can run by hand in an afternoon. It produces three numbers: your named rate, your own-page share and a ranked gap list of the pages cited where you are missing.
You also get the results of crawler, pricing and llms.txt checks we ran on well-known SaaS (software as a service) websites, and a table that maps each gap to a fix, in order of what to fix first.
What does AEO for SaaS involve, and where should you start?
AEO for SaaS is making ChatGPT, Perplexity and Google AI Mode name your product, and cite a page that supports it, when a buyer asks for a category shortlist. Start with a source-gap audit: count which pages those engines cite for your category, and whether you appear on them.
AEO, GEO and SEO. SEO ranks your pages in search results to earn clicks. AEO, as we use the term, aims to get your product named and cited inside AI answers. Many teams use generative engine optimization (GEO) for the same work; see GEO vs SEO for what changes in practice.
Two terms keep the count honest:
Definition
Definition
An answer can name you without linking you, or link your page without naming you. Log both, because they call for different fixes.
Which pages you don't own shape a software shortlist?
Review sites and community pages are among the most-cited sources, and buyers weight review sites heavily. Two datasets point the same way, though each has limits.
On the buyer side, G2's 2026 Buyer Behavior Report found review sites (38%) narrowly ahead of AI chatbots (37%) as the top influence on shortlists. In G2's AI Search Insight Report, a March 2026 survey of 1,076 decision-makers, nearly 7 in 10 said a chatbot led them to a different vendor than they expected. 33% bought from a vendor they had not heard of before, and 45% called review-site citations the most confidence-inspiring signal in an AI answer. G2 runs a software review site, so weigh the review-site finding with that in mind.
On the engine side, a study published at trellner.com, whose publisher is not identified, sent 380 software category prompts to Perplexity's sonar and sonar-pro models on September 2, 2026: 760 calls, 7,534 citations across 2,055 domains. The two most-cited domains were g2.com (291 citations) and reddit.com (261). A vendor's own marketing blog was third, with 194 citations, and three sites had published 215,128 generated "best software" guides. Many cited sources are low-traffic sites: 59.8% sat outside the 100,000 most-visited websites.
The limits matter. That test covers one engine on one day, with category prompts the researchers chose. Engines also differ from each other. In our October 2026 research at Citation Research, across 150 questions in ten topics, including software for business, asked mainly from India, Google AI Mode cited a community platform (Reddit, YouTube or Quora) in 117 answers and ChatGPT in 3. Read it as a sign that engines differ, not as a US software benchmark.
Our inference: because review sites and third-party lists are among the most-cited sources, check for a gap there before rewriting your own pages. You also cannot assume one engine's sources predict another's, so audit each engine separately. And some cited "best of" pages are generated filler, so a page appearing in your log is not automatically worth pursuing.

Is your own site blocking AI crawlers or hiding its prices?
Check three things: robots.txt, a network-level block, and plain-text prices. In our November 17, 2026 fetch of 60 SaaS sites, most AI crawlers were not blocked in robots.txt, but about 3 in 10 pricing pages showed few prices in plain HTML:
| Check | Result |
|---|---|
| GPTBot blocked in robots.txt | 3 of 55 readable files |
| OAI-SearchBot, ChatGPT-User and PerplexityBot blocked in robots.txt | 1 of 55 each |
| ClaudeBot blocked in robots.txt | 2 of 55 |
| Pricing pages with fewer than three dollar figures in plain HTML | 13 of 43 that loaded (9 showed none) |
| Sites with a working llms.txt | 37 of 59 checked |
These checks read robots.txt and raw HTML only, so they cannot see edge-level blocks or prices loaded by script, and some pages without prices may be sales-led by design. The domain list and full limits are in "How we ran the checks" below.
How we counted:
- Crawler access. A robots.txt file tells automated crawlers which paths they may fetch. We fetched it for 60 SaaS domains, and 5 returned no readable file. We counted a crawler as blocked when its own group, or the all-crawlers group if it had none, held a blanket
Disallow: /. Result: GPTBot was blocked on 3 of 55; OAI-SearchBot, ChatGPT-User and PerplexityBot on 1 of 55 each; ClaudeBot on 2 of 55. - Pricing visibility. We requested
/pricingon 53 SaaS domains with a plain HTTP fetch, which does not run scripts. 43 returned a page and 10 did not. Of the 43, 30 showed three or more dollar figures in the raw HTML. 13 showed fewer than three, and 9 of those showed none. - llms.txt. We requested
/llms.txton 60 domains. A site counted as having one if it returned HTTP 200 with plain text or markdown content, not an HTML error or redirect page. Of the 59 that returned a usable answer, 37 did (wix.com is excluded; see "How we ran the checks").
We chose the 60 domains by hand as well-known SaaS vendors across CRM, collaboration, marketing, security, developer and finance tools, so this is not a random sample. They were hubspot.com, salesforce.com, slack.com, asana.com, monday.com, notion.so, atlassian.com, zendesk.com, intercom.com, mailchimp.com, shopify.com, stripe.com, zoom.us, dropbox.com, box.com, airtable.com, clickup.com, trello.com, freshworks.com, pipedrive.com, semrush.com, ahrefs.com, canva.com, figma.com, miro.com, calendly.com, typeform.com, webflow.com, squarespace.com, wix.com, docusign.com, okta.com, datadog.com, twilio.com, sendgrid.com, klaviyo.com, gong.io, outreach.io, apollo.io, zoominfo.com, workday.com, servicenow.com, snowflake.com, hotjar.com, loom.com, grammarly.com, surveymonkey.com, basecamp.com, linear.app, smartsheet.com, quickbooks.intuit.com, xero.com, gusto.com, brex.com, ramp.com, mixpanel.com, amplitude.com, segment.com, pagerduty.com and sentry.io.
The fetch identified itself as Mozilla/5.0 (compatible; research-check). The 5 domains with no readable robots.txt were freshworks.com, servicenow.com, quickbooks.intuit.com, gusto.com and brex.com. The pricing check covered the same list minus stripe.com, wix.com, sendgrid.com, servicenow.com, quickbooks.intuit.com, brex.com and segment.com. Any pricing response other than HTTP 200 (404, 403 or a 308 redirect) counted as a page that did not load. On the llms.txt check, wix.com answered with an HTTP 429 (too many requests) on recheck and was not counted.
All three checks have limits. A robots.txt file is not the only way to block a crawler, so a clean file is not proof of access, and our 60-site robots.txt count cannot see edge-level blocks, meaning blocks applied at the network layer before a request reaches the site. A page with no visible prices may be sales-led, or may load prices by script, and the check cannot tell which. The dollar pattern also catches only US dollar prices.
Run these on your own site in about ten minutes:
- Open
yoursite.com/robots.txt. Fail: a blanketDisallow: /under GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot or Googlebot, or under the all-crawlers group (User-agent: *). GPTBot, OAI-SearchBot and ChatGPT-User are OpenAI's crawlers, PerplexityBot is Perplexity's, ClaudeBot is Anthropic's and Googlebot is Google's; blocking Googlebot removes the site from Google's search-based answers. Pass: none.
From September 15, 2026, Cloudflare's new defaults block Training and Agent bots on pages that display ads for new domains onboarding to Cloudflare, unless the site owner opted out. Search bots stay allowed by default. The block is applied at the network level, so it does not show in your robots.txt (Cloudflare).
- Run
curl -s https://yoursite.com/pricing | grep -o '\$[0-9][0-9,.]*' | wc -l. curl is a command-line tool that fetches a page without running scripts, so the result shows the raw HTML, the page source before any script runs. Pass: 3 or more. Fail: fewer than 3, meaning a plain fetch cannot see your plan prices. - Run our free Cloudflare AI crawler check. Fail: any crawler comes back blocked. Pass: none blocked.
If you publish prices, put them in the page text. Our recommendation, since an answer to "how much does this cost" can only quote what the engine can read.
llms.txt is a plain-text map of key pages for AI systems. 37 of the 59 sites we could check returned a real one. Whether it helps citations is a separate test, covered in our llms.txt guide.
How do you run a source-gap audit for your SaaS category?
Run 15 buyer prompts across ChatGPT, Perplexity and Google AI Mode, log every brand named and every source cited, then compute three numbers. One pass is 15 prompts times 3 engines, which is 45 answers. This is a sample, not a census. For a larger sample, see our AI visibility audit guide, which uses a 450-answer sheet.
Which 15 prompts should you run?
Use three prompts in each of five types, worded the way your buyers talk. Pull the phrasing from sales calls and support tickets. Our prompt research guide covers choosing prompts worth tracking.

| Type | Templates |
|---|---|
| Category | "best [category] software for [segment]"; "top [category] tools for [company size]"; "which [category] software should a [role] buy" |
| Alternatives | "[top competitor] alternatives"; "alternatives to [competitor] for [segment]"; "cheaper than [competitor]" |
| Versus | "[you] vs [competitor]"; "[you] vs [competitor] for [use case]"; "is [you] worth it" |
| Use case | "how do I [job] with software"; "software to [job] without [pain]"; "[job] tool that integrates with [platform]" |
| Pricing | "how much does [category] software cost"; "[competitor] pricing"; "[you] pricing" |
What do you log for each answer?
One row per answer, with these columns: prompt, engine, run number, brands named in order, you named (Y/N), each cited URL, its domain, and its source type. The source types are own site, review site, "best of" list, community, competitor site and other.
Answers change between runs, so run each prompt more than once per engine. Our AI citation tracking guide covers how many answers make a number real. One pass is 45 answers; we recommend three passes (135 answers) for a first read (our recommendation).
Which three numbers do you calculate?
Work out these per engine:
- Named rate = answers that name you ÷ all answers.
- Own-page share = citations to your domain ÷ all citations logged.
- Gap list = domains cited in 3 or more answers whose text does not name you, ranked by count. The threshold of 3 is our recommendation, to keep one-off sources out.
Compare your named rate with your closest competitor's from the same log. A gap domain is worth pursuing only if real buyers would read it and it has a way to reach the author or submit a listing.
What should you fix first once the log shows your gaps?
First fix any crawler block, since nothing else matters if engines cannot fetch your pages. Then work the pattern your log shows most often. Each row pairs a test you can read off the log with the fix.
| What the log shows | Test | Fix |
|---|---|---|
| Crawlers blocked | Blanket | Remove the block and re-run the sheet |
| Competitors named, cited pages are lists you are not on | A list domain appears in 3 or more answers and you are absent | Pitch the author a data point or ask for inclusion. See third-party brand mentions and generative engine optimization (GEO) |
| Review-site pages dominate the gap list | A review domain is in 3 or more answers and you have no profile, fewer than 10 reviews, or no category listing | Complete the profile and ask customers for reviews |
| Community pages (Reddit, YouTube, Quora) dominate the gap list | A community domain is in 3 or more answers and you are absent | Answer the cited threads where the community's rules allow it, and publish a demo or walkthrough video |
Competitor sites are cited on "[you] vs [competitor]" prompts | A competitor-owned page is cited on Versus prompts and your own comparison page is not | Publish your own comparison page (see the next section) |
| Your pricing or comparison page is never cited | Own-page share is 0 on pricing prompts and the curl check shows fewer than 3 dollar figures | Put plan prices and a comparison table in plain page text |
| You are named but never cited | Named rate above 0, own-page share 0 | Publish a page that states the claim with data the engine can quote |
Which of your own SaaS pages should you fix next?
Once the log shows which gaps you can close, fix your own pages in this order (our recommendation):
- Pricing page. Put plan names and prices in page text. In our check, 13 of 43 pricing pages showed fewer than three dollar figures in plain HTML.
- Comparison and alternatives pages. Write one page for each competitor your log names instead of you. Add a table of features, pricing and best-fit buyer, and say where the competitor is the better choice.
- Integration pages. Write one page for each platform buyers name in "[job] tool that integrates with [platform]" prompts.
- One product description everywhere. Use the same product name, category and one-line description on your site, your review-site profiles and LinkedIn.
- Structured data. Add SoftwareApplication and Organization markup that matches the visible page text. See schema markup for AI search.
Then re-run the sheet and check whether your named rate and own-page share moved.
How do you tie AI answers to trials?
AI answers rarely produce a last-click signup, so measure in four ways:
- AI referral sessions. In GA4, segment sessions from ChatGPT, Perplexity and Gemini. See how to track AI traffic in GA4 for the channel group and regex.
- Assisted conversions. Compare the AI-referred segment with trial signups that happened later through another channel.
- Self-reported source. Add a "how did you hear about us" field to the signup form.
- Branded search. Watch branded queries in Google Search Console for growth after your named rate rises.
Our Analytics feature ties cited pages to traffic.
How do we find and close source gaps at PageLens.ai?
We do it in three steps, and the first is free. Our Free AI Visibility Audit needs only your domain and no signup or credit card. In about 30 seconds it returns a 0 to 100 visibility score, the language each model uses about your brand, the competitors AI recommends instead, and the citation sources where those competitors appear. That is the fast version of the gap list.
Second, our prompt research finds the buyer prompts in your category, so the sheet above runs on questions buyers actually ask.
Third, our outreach agent works the gap list. It finds the publications and creators AI already cites in your category, ranked by citation frequency, researches each target's recent work and writes a tailored pitch. It queues one pitch a day for your approval before anything is sent, and it tracks replies and the citations that follow.
Plans, as of November 2026:
- Launch: $299 a month. Tracks ChatGPT, Google AI Mode and Perplexity.
- Growth: $699 a month. Adds Gemini and Grok, and includes the outreach agent.
- Enterprise: from $1,499 a month. Seven engines, and all Growth features.
Every plan has unlimited team seats and a 7-day free trial.
If you want your gap list read with someone who does this daily, Book a demo.
See which pages AI cites instead of yours
Frequently asked questions
Sources
- 1.G2 2026 Buyer Behavior Report, G2, July 22, 2026
- 2.G2 2026 AI Search Insight Report, G2, March 2026
- 3.Study of Perplexity software recommendations, trellner.com (publisher not identified), September 2, 2026
- 4.Citation Research Study 007: Reddit, YouTube and Quora in AI citations, Citation Research (PageLens.ai's research lab), October 2026
- 5.Cloudflare's announcement on AI crawler options, Cloudflare, September 15, 2026
- 6.PageLens.ai's own check of 60 SaaS domains' robots.txt and llms.txt files and 53 domains' /pricing pages, by plain HTTP fetch, PageLens.ai, November 17, 2026


