Own the shortlist, then measure the signups it sends
B2B software evaluation now routinely starts with an assistant: what are the options, how do they compare, which suits this situation. That produces a shortlist of three or four vendors before anyone visits a website or fills in a form. Being off it is expensive and almost entirely invisible in your funnel reporting.
For a SaaS marketing team the work is unusually concentrated. A small set of buying prompts, a small set of cited sources per category, and a few surfaces you control, mainly documentation, comparison pages and review profiles, decide most of the outcome. This page covers how to map those prompts, where the shortlist is actually decided, and how to connect the channel to signups and pipeline.
Why SaaS teams track this
The shortlist is the funnel
When a buyer asks for options in your category, the assistant names a handful of vendors with a line of reasoning each, and the evaluation starts from that list. If you are not on it there is no impression to lose, no session to attribute and no data point showing it happened. The only way to know is to ask the same questions on a schedule and record who is named. The Visibility Tracker at /features/brand-visibility-tracker does that per engine, so the absence becomes a number rather than a suspicion.
Review platforms carry unusual weight
Software recommendations lean on G2, Capterra, TrustRadius and their equivalents more than almost any other category leans on any source, because those sites already structure the comparison an assistant is trying to make. Profile completeness, category placement, review recency and the wording of reviews all feed what the assistant says about you. That makes review operations an AI-visibility lever, and it is usually owned by a team that has never heard the term.
Comparison content is the battleground
"X vs Y" and "alternatives to X" questions sit closest to the purchase decision, and the sources that answer them are a small, identifiable set: a few comparison posts, a review site or two, a community thread. Whether your own comparison pages are cited depends on whether they are honest and specific enough to be treated as a source rather than an advert. Citation Intelligence at /features/citation-intelligence shows which pages win those prompts today.
Documentation is a citation surface
Technical and implementation questions are answered from documentation, and a vendor whose docs are open, server-rendered and well structured becomes the source of the correct answer for its own category. Docs are frequently the weakest surface: behind a login, rendered client-side by a docs framework, or blocked by a broad bot rule. Fixing that is cheap and puts your product name in answers that competitor marketing pages will never reach.
How it works for SaaS
From absent to shortlisted in a software category, and from shortlisted to a number in the pipeline report.
- 1
Map the buying prompts
List the questions a buyer asks on the way to a decision: the category question ("best tools for X"), the use-case questions ("X for a team that needs Y"), the comparison questions ("A vs B"), the alternatives questions ("alternatives to A") and the implementation questions your docs should answer. Pull real phrasing from sales calls, support tickets and community threads rather than from keyword tools. Thirty to fifty prompts per category is enough to start. Prompt Intelligence at /features/prompt-intelligence surfaces the phrasing buyers actually use.
- 2
Baseline the shortlist
Run the prompt set against ChatGPT, Perplexity, Google AI Mode, Gemini and Copilot on a schedule and record who is named, in what order, with what reasoning, and which sources are cited. Do this before changing anything so the trend line has a start. The cited-source list is usually more actionable than the ranking, because it tells you where the answer is being decided and whether you are present there at all.
- 3
Fix access and rendering
Confirm OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User and ClaudeBot can fetch product pages, pricing, comparison pages and documentation, and that bot management does not challenge them. Fetch each template with a non-JavaScript client and confirm the body and JSON-LD are present. Documentation frameworks and pricing pages are the usual failures. Check the served robots.txt with /tools/robots-txt-validator and keep training-crawler rules separate from retrieval-agent rules.
- 4
Make comparison and pricing pages citable
Write comparison pages that state specific, verifiable differences, name the cases where the competitor is the better choice, and keep the structure answer-first. Publish current pricing in plain text with Offer schema rather than only in an interactive calculator. An assistant that cannot read your pricing will repeat a third party's stale version of it. Add FAQPage markup that matches visible questions, and validate with /tools/json-ld-validator.
- 5
Work the cited sources
For each source on the cited-source list, decide whether you should be present, more complete or more recent there. Review profiles: complete every field, keep category placement accurate, and run a steady review programme so recency holds. Comparison posts: pitch the authors with specific facts. Community threads: participate honestly where the product is being discussed. Reddit monitoring at /reddit-monitoring surfaces the threads being cited. See /solutions/off-site-authority for the full method.
- 6
Connect to pipeline
Recover AI-assisted sessions from Direct with the AI Traffic Decoder at /features/ai-traffic-decoder, tag them in your analytics, and follow them through to signup and opportunity. Add a "how did you hear about us" field with an assistant option to the signup form as a second signal. Report shortlist presence, cited-source coverage and AI-assisted signups together, and treat the signup number as a floor because most of the influence never produces a click.
SaaS use cases
Category shortlist tracking
Whether you are named when a buyer asks an assistant for options in your category, per engine, over time. This is the headline number for the channel and the one that tells a marketing lead whether the programme is working.
Competitive comparison monitoring
How assistants characterise you against named competitors, which differences they mention, and which sources they cite for the comparison. It catches stale or wrong claims early, and it shows whether your own comparison pages are being treated as a source.
Documentation visibility
Whether implementation and troubleshooting questions in your category are answered from your docs. Docs are cited heavily in technical categories and are frequently blocked, behind a login or rendered client-side, which makes this one of the cheapest fixes available.
Pricing accuracy
What assistants say you cost, compared with the truth. Models repeat stale pricing from old reviews and comparison posts for a long time. Publishing current pricing in readable text with schema, and correcting the sources that carry the old version, fixes it.
Review platform coverage
Which review sites are cited for your category and how complete and recent your profile is on each. Review operations is an AI-visibility lever in software, and the cited-source list tells you which platforms matter for your category specifically.
What differs per engine for B2B software
The buyers differ by engine as much as the mechanics do, so track all of them and weight by where your signups come from.
ChatGPT
Search-mode answers fetch pages live through OAI-SearchBot and ChatGPT-User, so access and server rendering for docs, pricing and comparison pages are the first check. Category answers lean on review platforms and comparison posts; implementation answers lean on documentation. Citations show as links, so citation share is directly measurable. Description accuracy varies by session, which is why a prompt set is tracked rather than a screenshot.
Perplexity
Perplexity is the engine technical evaluators reach for, and it shows sources prominently, which makes it the fastest feedback loop for whether documentation and comparison changes registered. PerplexityBot indexes and Perplexity-User fetches at query time. It favours recently updated, specific pages, so accurate dateModified values and precise feature detail help more here than anywhere.
Google AI Overviews and AI Mode
Both draw on the Google index via Googlebot, so existing crawlability carries over and Google-Extended is not the control. AI Mode splits a category question into sub-questions and cites pages that answer each directly, which rewards answer-first comparison and pricing pages. Search Console still shows what Google has, but the overview can cite pages that never ranked on the classic page.
Gemini
Gemini grounds on Google Search and is the default assistant inside Google Workspace, so it reaches buyers in Workspace-centric companies. Google-Extended governs training and grounding use beyond search, so blocking it for policy reasons will thin coverage here. A consistent Organization record with sameAs, and clear product entity naming, affect how Gemini describes you against competitors.
Microsoft Copilot
Copilot grounds on the Bing index and is the default in Microsoft-centric enterprises, which for many B2B categories is the buyer. Bingbot access and Bing Webmaster Tools coverage are the levers, and sites tuned only for Google often have gaps. IndexNow speeds pickup of updated docs and pricing. Citation display resembles ChatGPT in search mode.
The numbers SaaS teams report
Shortlist presence is the headline; documentation citation share is the early win; signups are the floor.
- Shortlist presence
- The share of category and use-case prompts where you are named, per engine. It is the headline reach metric. Movement is slow and depends on off-site sources, so compare month to month on a fixed prompt set and read the trend rather than any single run.
- Comparison win rate
- On "X vs you" and "alternatives to X" prompts, how often you are named and whether the reasoning favours you, is neutral, or repeats a competitor's framing. Read the reasoning text, not just the count; it tells you which claims and which sources are shaping it.
- Documentation citation share
- The share of implementation and troubleshooting prompts in your category answered from your docs. It moves quickly after access and rendering fixes, which makes it a good early win to report while the slower category work continues.
- AI-assisted signups and pipeline
- Sessions recovered from Direct and attributed to answer engines, followed through to signup and opportunity, plus self-reported attribution. Present it as a floor: the influence that never produces a click is real but unmeasured. It is the number finance will ask for.
Mistakes SaaS teams make on the first pass
Mistake · Gating documentation behind a login or rendering it client-side.
Fix · Serve public docs server-rendered and open to retrieval agents. Docs are the surface that puts your product name into technical answers, and a docs framework that hydrates on the client hides it from every fetch.
Mistake · Publishing pricing only in an interactive calculator or behind "contact sales".
Fix · State current plans and prices in plain text with Offer schema, even alongside a calculator. An assistant that cannot read your pricing repeats a third party's stale version instead.
Mistake · Writing comparison pages that only list where you win.
Fix · State specific differences and name the cases where the competitor is the better fit. Honest, verifiable comparisons get cited; pages that read as adverts are passed over for third-party comparisons.
Mistake · Treating review platforms as a sales concern rather than a visibility one.
Fix · Complete every profile field, keep category placement accurate and run a steady review programme. Review sites are among the most cited sources in software categories, and recency matters.
A worked example: a mid-market workflow automation platform
A hypothetical mid-market workflow automation platform with strong organic rankings notices that assistants asked for options in its category name two larger competitors and a newer entrant, and never it. The baseline across forty prompts confirms it: shortlist presence is near zero on ChatGPT and Google AI Mode, slightly better on Perplexity, and the cited sources are two review platforms, one comparison blog and a handful of community threads.
The technical audit finds the documentation site, built on a client-rendered docs framework, returns an empty shell to a non-JavaScript fetch, and the pricing page shows plans only through a calculator. The docs are moved to static rendering, a plain-text pricing section with Offer schema is added above the calculator, and the CDN allow list is extended to the retrieval agents.
On the review platforms the profile is complete but sits in the wrong category and has no reviews from the last year. Marketing corrects the placement and starts a review programme tied to onboarding milestones. The comparison blog's post predates the current positioning; the team sends the author a factual update and publishes its own comparison pages that name the situations where each competitor is the better choice.
Documentation citation share moves first, within weeks, as implementation questions start being answered from the docs. Shortlist presence follows on Perplexity, then Google AI Mode, with the review platform cited as the reason. AI-assisted signups recovered from Direct are reported monthly as a floor next to the shortlist number, and the cited-source list becomes a standing agenda item for the content and review teams.
Frequently asked questions
Why does the assistant recommend competitors but not us?+
Usually because the sources it cites do not mention you, not because it prefers them. Pull the cited-source list for the category question and check which of those sources cover you at all. That gap is normally the whole story, and it is fixable through review profiles, comparison content and community presence.
Should our documentation be open to AI crawlers?+
For most B2B software, yes. Technical answers cite documentation heavily, and being the source of the correct implementation answer is a strong position. Public documentation should be crawlable by the retrieval agents, server-rendered, and not behind a login. Keep genuinely private material separate rather than gating everything.
How much do review sites matter?+
A great deal in software categories; they are among the most frequently cited sources for category and comparison questions. Profile completeness, category placement, review recency and the specific wording of reviews all feed what the assistant says. Treat review operations as part of the AI-visibility programme.
Can we measure this against pipeline?+
Partly. AI-assisted sessions can be recovered from Direct and connected to signups and opportunities, and a self-reported attribution field adds a second signal. Attribution is not complete, because much of the influence happens in a conversation you never see, but it is far better than the nothing most teams report today.
Do our own comparison pages get cited?+
They can, if they are specific, verifiable and honest about where the competitor wins. Pages that read as adverts are skipped in favour of third-party comparisons. Answer-first structure, real feature-by-feature detail and current pricing make the difference, and citation tracking will show whether it worked.
Which engines matter most for B2B?+
It depends on your buyers, which is why you measure rather than guess. ChatGPT and Google AI Mode reach the widest audience, Perplexity is used heavily by technical evaluators and shows sources clearly, Copilot matters in Microsoft-centric enterprises, and Gemini in Google Workspace shops. Track all of them and weight by where your signups come from.
How do we handle a wrong claim in an AI answer?+
Trace the source. The answer will cite or paraphrase a page, and that page is where the correction goes: your own pricing page, a review profile, a comparison post or a community thread. Fix or contact the source, then re-run the prompt on a schedule until the answer changes. Correcting the assistant directly does not persist.
How long before we see movement?+
Access and rendering fixes register within days on engines that fetch live. Comparison and documentation content moves over weeks as it is fetched and cited. Review and community work is the slowest because it depends on third parties. Baseline first, report monthly, and expect Perplexity to move before ChatGPT.
Explore the features behind this solution
Go deeper
Are you on the AI shortlist?
See who gets named when a buyer asks an assistant about your category, and which sources decide it.