Perplexity optimization for B2B software: how to get cited

Perplexity describes itself as an answer engine that searches the web in real time and condenses what it finds into one answer with numbered citations to the original sources. Unlike a model answering from memory, every cited claim points to a URL, which makes Perplexity the easiest assistant to diagnose: you can see exactly which pages it used.

Perplexity's research team explained the retrieval side in September 2025. Its index tracks more than 200 billion unique URLs. Pages are split into self-contained spans that are retrieved and ranked individually, so the system can hand the model the most atomic useful units. Retrieval mixes lexical and semantic matching, a prefilter removes non-responsive or stale content, and progressively stronger rankers, ending with cross-encoders, score both documents and spans.

Two consequences follow for B2B software. Your answer has to survive being read as a short passage without the rest of the page, and stale pages can be filtered before ranking even starts. A pricing paragraph that only makes sense next to a toggle, or an integration list that only loads in a script, is at a disadvantage.

Perplexity runs two agents, and they need different handling.

Perplexity publishes its IP ranges as JSON files and recommends firewall rules that match both the user agent and the source IP, updated automatically. Security teams at software companies often block unknown bots by default, so check your bot protection logs, not only robots.txt.

Span-level retrieval changes how we write product and comparison pages. Each section should stand on its own.

1. Open each section with a direct answer to one question, in one or two sentences that name the product and the subject.
2. Use specific values with units and dates: API rate limits, supported regions, data retention, seats included, implementation weeks.
3. Name integrations, standards and certifications in text, not only as logos.
4. Keep pricing, plans and limits in HTML text on the page, not in images, PDFs or scripts that load on click.
5. Repeat the product name instead of "it" or "our platform" at the start of key passages, so a span still makes sense alone.
6. Show a visible last-updated date on pages with facts that change, and update it only when the facts change.

When we run software buyer prompts in Perplexity, the answers lean on a predictable set of page types. Build or fix these first.

Perplexity cites what ranks in its index for the question, and that is often not your site: review platforms, analyst and media coverage, partner marketplaces, community threads and other vendors' comparison pages. G2's 2026 survey of 1,076 software buyers, which is vendor research from a review platform, found 71% use AI chatbots at some point in their research and that buyers rank review-site citations as their top trust signal in AI answers.

Record the cited domains for every prompt in your set, then decide which ones you can realistically influence: your marketplace listings, partner directory pages, review profiles and the facts in third-party comparisons. Our AI shortlist authority work covers that off-site program.

Perplexity also matters inside companies. Its Enterprise Pro product can search the public web and a company's internal files in one query, so an evaluation team may be reading your public pages next to their own requirements document.

Perplexity is the simplest engine to measure because every answer shows its citations. Run a fixed prompt set monthly, record whether you are named, whether you are cited and which URL is cited, and note the competitors and domains that appear instead.

In analytics, segment referrals from perplexity.ai by landing page, and where you use account identification, by company. Our guide to measuring AI search visibility in GA4 shows the channel setup. Then follow those sessions into the CRM: AI referrals are usually small in volume but come from buyers who have already done their research.

Most are easy to fix once someone looks.

• Blocking PerplexityBot through a default bot protection rule nobody reviewed.
• Integration and security facts only in sales decks and gated PDFs.
• Pricing pages that render plans with client-side scripts and show nothing in the HTML.
• Comparison pages that claim to win every row, which reads as marketing and is less likely to be quoted.
• Old pages that still describe retired plans or features and keep getting cited.

Begin with a prompt set of 30 to 50 buyer questions, a crawler access check and a review of your integration, pricing and security pages. Most software companies see the biggest gains from making those three page types specific and public.

Our generative engine optimization service covers Perplexity alongside ChatGPT, Google AI Mode and Copilot.

For US software companies, our page on GEO in the United States explains how we work with teams there.

To see what Perplexity cites for your category today, talk to our team.