In the TrustRadius 2026 B2B Buying Disconnect report, 63% of the 1,862 technology buyers surveyed had used AI during their purchase, and 83% shortlisted three products or fewer. They also check what AI tells them. Of the buyers who used AI, 94% said they fact-check its responses at least some of the time.
SaaS AEO is answer engine optimization for software companies, meaning, the work of getting your product recommended, and described accurately, in the answers ChatGPT, Gemini, Claude, Perplexity, and Google's AI features give buyers. The survey numbers above split it into two jobs. You need to make a shortlist that's only a few names long, and you need what AI says about your pricing, integrations, and features to hold up when a buyer checks it.
Much of AEO works the same in any industry, and our AEO Guide covers it in depth. This guide focuses on what's specific to software. including the prompts buyers use to compare tools, the pages AI treats as your source of truth, and the review sites and listicles it leans on.
How AI engines decide which software to recommend
When a buyer asks which project management tool fits a 40-person agency, an AI engine will often search the web, read a set of pages, and write an answer from what those pages say. Not every engine does it every time, though.
In our study of Claude and Claude Code, Claude searched in 93% of responses, while Claude Code searched in only 13% and answered from what the model already knew the rest of the time. So what you publish today shapes the engines that search, and what was on the web when a model was trained shapes the engines that answer from memory.
The pages engines read come from a mix of sources. Across 11.84 billion citations in our citation sources study, brand sites, meaning a company's own website, docs, and blog, made up roughly 57% of citations, though the mix varied by engine. Brand sites were 47% of ChatGPT's citations and 69% of Gemini's, and earned media made up 30% of ChatGPT's. For software queries specifically, G2 was the fourth most-cited source on ChatGPT.
The question influences the answer as much as the sources do. In chapter three of our AEO Guide, we sort prompts into six types, and four of them have the most sway in software:
- Discovery prompts ask what's out there, like "best expense management software for startups."
- Competitor prompts put you next to a rival, like "Asana vs. Monday.com" or "alternatives to Jira."
- Validation prompts test a choice the buyer is leaning toward, like "is Notion good for a 200-person company?"
- Accuracy prompts check facts about important factors, such as pricing, features, integrations, and specs.
Build your prompt set from how buyers shop for software
Your prompt set is the list of questions you track across engines, and it determines what you see. If you only track "best [category] software," you'll miss most of what buyers ask. In our study of ChatGPT conversations, only 21.8% of B2B software and SaaS conversations were commercial. Another 37.4% were informational, people trying to understand a problem, and 40.8% were generative, people asking AI to produce something.
To build a set that reflects your buyers, write prompts from three inputs: the buyer’s persona, their segment, and their use case. “Best CRM” then becomes, for example, “best CRM for a RevOps lead at a 50-person SaaS company moving off spreadsheets,” which is much closer to how a real buyer describes the problem.
Before you lock the set, check the phrasing against real demand. Profound's Prompt Volumes shows what people ask AI engines about your category, with volume over time and intent, so you're tracking questions people ask instead of questions your team assumes they ask. Tag prompts by product line as you go. If you sell more than one product, each one needs its own view of visibility, because a strong number for your flagship can hide weak results for the product you launched last quarter.
You don't need months to get started. Statsig's marketing lead, Skye Scofield, used Profound to set up the core prompts his team cared about in under two hours and reviewed the answers and cited sources the next day. His team then published content on product analytics subcategories and comparisons with legacy tools, and nearly doubled its visibility across answer engines in key categories in under a week.
Fix the pages AI treats as your source of truth
Some pages on your site carry more weight than others. Treat prominent pages like your homepage, pricing, about, and comparison pages as their own category, because they get heavy crawl attention and are often where AI looks for basic facts about a company. For software, docs and integration pages belong on that list too.
Pricing
Pricing is the claim AI gets wrong most often. In our study of 158,000+ FactCheck claims, pricing and billing made up 12% of the claims evaluated but 24% of the claims found inaccurate. Among brands with at least ten pricing claims evaluated, pricing was the theme with the most inaccuracies for 66% of them. And 54% of brands had at least one inaccurate claim that cited their own content.
That last number means the solution often starts on your own site. Old plan names in a help article or a pricing FAQ that predates your last packaging change can end up in an answer long after the pricing page changed. Put current pricing in text, not only in images, on one page, keep plan names consistent everywhere they appear, and update supporting pages the same day pricing changes. Buyers want it, too. TrustRadius’ study also found that transparent pricing has been buyers' top request of vendors for four years running.
Documentation
Docs used to be written for customers. Now coding agents read them too. In the same Claude Code study, nearly three-quarters of the visits we observed from Claude Code's agent went to documentation, informational, and pricing pages, compared with 5% for Claude's. Our recommendation is to make documentation and pricing pages accurate and machine-readable, with specific claims, question-shaped headings, and answers placed first.
MongoDB built this into its process. Fiona Erickson's organic acquisition team used Profound's API to compare AI answers to branded questions against MongoDB's own verified answers. Inaccurate citations went to the content team to fix at the source. MongoDB now sees 90%+ accuracy in AI responses, five times the citations in AI search, and 50% higher answer engine visibility.
Integrations
In B2B, what you connect to is often what decides the deal. When a buyer asks whether your tool works with the systems they already run, the engine needs a page that answers. Give each important integration its own page that says what it does, what it doesn't do, and how setup works. A single logo wall gives the engine nothing to quote.
Comparison and alternatives pages
Your own comparison pages are where you get to make your case on competitor prompts. Write each page for a buyer who’s about to check it. Include:
- Specifics a buyer can verify. Pricing, plan limits, integrations, and setup time for both products, each with the date you last checked it.
- Where the competitor is stronger. Naming it makes the rest of the page more believable, and buyers who aren’t a fit can rule themselves out before a sales call.
- What switching involves. Migration steps, import tools, and how long a move usually takes, since a buyer comparing two tools might want to know what changing costs them.
Then keep the pages current. A competitor’s pricing change makes your page wrong about them, and an engine can repeat that error in an answer that mentions you both. Review each page when a competitor ships something significant, and at least once a quarter. In Profound, you can track your competitor prompts to see which comparisons leave you out and which pages the engines cite for them, including whether a competitor’s own comparison page is the source behind the answer.
Win the third-party sources buyers and engines trust
Your site only covers part of what the engines read. The rest comes from review sites, publishers, partners, and communities, and buyers check many of the same places. In the TrustRadius survey, 74% of buyers used reviews to inform their purchase.
Four kinds of third-party sources are worth your attention:
- Review sites. G2, Capterra, and TrustRadius are heavily cited by AI models, so what they say about you is often close to what AI says. AI also learns from the language reviewers use. Recent, specific reviews give the engines more to work with than a star rating does.
- Listicles and comparison articles. They’re the most common off-page tactic in B2B, and one that works, though avoid the temptation to over-index on one format.
- Communities. Our Reddit research found that engines tend to draw on three to five key subreddits per prompt.
Third-party pages also carry errors. In the inaccurate claims study, 51% of brands had at least one inaccurate claim that cited earned media, and pricing accounted for 35% of those. When you find one, the solution is outreach to the publisher with the correct information, the same way you'd fix your own page.
Treat every launch and pricing change as an AEO update
A launch, a new pricing tier, a renamed feature, or a rebrand can make yesterday’s accurate answer wrong. Engines that search the web keep finding the old details wherever they still live, like a help article with last year’s plan names, a partner page written two years ago, or a listicle nobody has updated. Engines that answer from memory lag further behind, as the Claude Code numbers earlier show, because they reflect what was on the web when the model was trained. You can’t control when a model is retrained, but you can make sure the current version is the one it finds everywhere.
That’s why AEO can’t solely be marketing’s responsibility. Give Product ownership of the docs, changelogs, and integration specs that describe what your product does, and put AEO on the launch checklist both teams use. The checklist has three stages:
- Before it ships. List every page that describes what’s changing. That includes your pricing page, docs, help center, and comparison pages, plus your review site profiles, partner listings, and the listicles that mention you. Add prompts for the new name, feature, or tier now, so you have a baseline to compare against.
- The day it ships. Update your own pages at the same time as the announcement, including older help articles that still use old names, and redirect pages that no longer apply. Update the review site profiles and partner listings you control on the same day.
- The weeks after. Check the prompts that cover the change. Where an answer is still wrong, find the page the engine cited. Fix it if it’s yours, and send the correct details to the publisher or partner if it isn’t.
In Profound, tags give your launch prompts their own view, FactCheck traces each wrong claim to the page it came from, and agents can draft the page updates and publisher outreach for your team to review.
Optro went through the biggest version of this, a full rebrand from AuditBoard. Ron Samson, its Sr. Growth Marketing Manager, set up one group of prompts to compare the old and new names over time and another to test persona- and industry-specific questions. His team used Profound agents to update 31 articles in a week, and its enterprise risk articles were cited within two days of publishing. Citation share on GRC prompts rose 70%.
Measure what AI search does for pipeline and signups
Much of what AI does for a software company never shows up as a click. For instance, a buyer might ask ChatGPT for project management tools, see your name on the shortlist, and a week later type your URL into the browser or search for you on Google. Your analytics record that visit as direct or branded traffic, and the answer that started it gets no credit.
In chapter six of our guide, Kyle Poyar of Growth Unhinged shows the size of that gap. "In my own data, LLMs account for <1% of new subscribers on a click basis," he says. "But 13.7% of new subscribers say they first heard about Growth Unhinged from an LLM." His advice is to measure self-reported attribution alongside click data. For a product-led company, that can be as simple as adding AI tools as an option in your signup survey. A sales-led team can ask the question in discovery calls and record the answer in the CRM.
Profound covers the layers you can measure directly. Answer Engine Insights tracks your visibility, citations, and share of voice by prompt, topic, and engine. Agent Analytics shows when AI crawlers visit your pages and how many people arrive from answer engines, and its GA4 integration brings in conversions and revenue. Read alongside self-reported attribution, those numbers give you a clearer view of the channel than either does alone.
Elevate your SaaS AEO program with Profound
We built Profound because buyers started asking AI engines what to buy, and marketers had no clear view of what those engines said back. So we started with the basics, including the prompts buyers ask, where you show up in the answers, which pages each engine cites, and what it gets wrong about your pricing and product.
That tells you what’s happening. For a software company, what’s happening changes with every release, pricing update, and new competitor, and the harder part is the work that follows: deciding which change will have the most impact, finding the pages to fix, and getting it done before the next release. We think AEO software now has to take on that work too, with agents doing the legwork and your team deciding what’s worth doing.
So we built AI Marketer (Aim). “It tells your team what changed, why it matters, and what to do about it,” our co-founder and CEO, James Cadwallader, explained when we launched it. Aim works in the background across Profound’s data and the context your team keeps in Context Manager, like your offerings, positioning, and upcoming launches. Every week, it brings a ranked list of opportunities with the data behind each, such as a pricing claim the engines still get wrong after your last packaging change, a competitor gaining on your comparison prompts, or rising demand for a use case you don’t cover yet.
When your team accepts an opportunity, Aim turns it into a project and builds the agents to do the work, whether that’s refreshing your docs, drafting a correction for a publisher, or adding FAQs to a comparison page. It tracks whether the answers change afterward, and nothing goes live without your approval.
See what Aim would put on your team’s list after your next release. Book a demo.
SaaS AEO FAQs
How is AEO different for SaaS than for other industries?
Mostly in which pages and sources matter, and how fast they go out of date. Software buyers ask detailed questions about pricing, integrations, and features, so docs, pricing pages, and integration pages carry more weight than they would for a consumer brand. Review sites like G2, Capterra, and TrustRadius, along with comparison listicles, are among the sources engines lean on most for software.
Software also changes constantly. Every release and pricing change can make an AI answer wrong, which is why SaaS AEO needs Product involved alongside marketing. And product-led companies need self-reported attribution, because so many AI-influenced signups arrive without a referral.
Do G2 and Capterra reviews affect AI recommendations?
Yes, because the engines cite review sites heavily when they answer software questions. G2 was the fourth most-cited source on ChatGPT for software queries in our 2025 analysis, and chapter two of our AEO Guide notes that what review sites say is often close to what AI says.
Keep your profiles accurate, especially pricing, features, and integrations, and ask happy customers for reviews regularly so new reviews keep coming in. Specific reviews help more than generic praise, since the language reviewers use is the language AI learns from.
How does Profound find what AI gets wrong about our pricing?
With FactCheck, which compares the claims AI answers make about you against your knowledge base and flags the inaccurate ones. For each wrong claim, it shows the source the engine relied on and the correct version. That source could be your own help center, a competitor’s comparison page, or a review site.
Can Profound show which docs and pricing pages AI crawlers visit?
Yes, through Agent Analytics. It reads your server logs instead of relying on a JavaScript tag, so it records AI crawlers and agents that never run a tag, and shows which pages they visit and how often. It also separates real bots from spoofed crawlers.
If your docs live on a separate subdomain or platform, connect that too. Agent Analytics works with providers like Cloudflare, Akamai, Fastly, AWS, Vercel, Netlify, and WordPress, and its GA4 integration adds AI-referred visits and conversions.
How does Profound track a product launch or rebrand?
With dedicated prompt groups and tags. Before launch, set up prompts for the new product, feature, or name, and tag them so they get their own view. Over time, you can compare them with the prompts for whatever they replace. Prompt Volumes shows whether people have started asking about the new name or feature.
After launch, FactCheck flags answers that still describe the old pricing or packaging. Agents can refresh the docs, FAQs, and comparison pages that need updating, with your team approving each change. Aim can also bring the gaps to you as opportunities, like a new feature the engines haven’t picked up yet.
Can Profound show which review sites and listicles the engines cite for our category?
Yes. Answer Engine Insights shows which domains and pages each engine cites for the prompts you track. You can see whether G2, Capterra, TrustRadius, or a particular listicle is shaping the answers in your category. Prompt Volumes also flags prompts where a competitor is cited and you aren’t.