We analyzed AI visibility for 20 B2B SaaS brands. Here's what we found.

Why we ran this scan

Building a platform for AI visibility means we need a clear picture of what AI visibility actually looks like in the wild — not in controlled demos, but across real production domains running real content. Before launch, we ran Rankwize's full monitoring and diagnosis pipeline across 20 B2B SaaS domains we selected for category diversity: project management, CRM, analytics, HR tech, security, developer tools, and marketing software.

These are companies with real SEO programs, real content teams, and real pages already indexed by AI engines. They're the typical Rankwize buyer profile. We wanted to understand what the citation landscape looks like before a team has done any AEO work — baseline state.

How we ran it

For each domain, Rankwize ran its full Prompt Discovery pipeline: domain fingerprinting, content library indexing, prompt generation across buyer-intent stages, four-factor scoring (demand, citation opportunity, content readiness, commercial value), and routing of the top-scoring intents into active monitoring. We then ran live API calls across all five platforms — ChatGPT, Google AI Mode, Google AI Overviews, Perplexity, and Copilot — and classified each response as cited, mentioned, or invisible.

For every invisible result, we ran full diagnosis: content library matching, gap classification into one of seven fix types, and recommendation generation. This is the same pipeline Rankwize runs in production for paying customers.

Finding 1: Roughly 6 in 10 gaps are fixable with content you already have

The most striking result: across all 20 domains, roughly 6 in 10 diagnosed gaps were classified as Update Existing — meaning the domain already had a page that matched the prompt, but that page needed to be improved, not replaced or supplemented with something new.

This contradicts the instinctive response to AI visibility gaps, which is to create more content. The data says most brands already have the pages they need. The pages are just not structured, evidenced, or positioned the way AI engines need to confidently cite them.

The remaining gaps split across the other six fix types — Add Evidence, Create Comparison, Create How-To, Create Diagnostic, Improve CTA, Offsite Seed — with no single type dominating outside of Update Existing. The exact distribution varied by domain vertical and content maturity, but the Update Existing pattern held consistently across the set.

Finding 2: Per-platform citation rates diverge significantly

Citation rates varied substantially by platform, and the pattern was consistent across domains. No brand in the set had uniform citation rates across all five platforms. The most common pattern was strong performance on one or two platforms and near-zero citation on the others.

Google AI Overviews was the most selective — the highest citation threshold, the narrowest acceptable content formats. Perplexity cited more broadly. ChatGPT fell between them. AI Mode and Copilot showed the most variance by domain, likely reflecting their different underlying models and retrieval strategies.

This matters for prioritization. A team that sees low overall citation rate and optimizes generically may be addressing the wrong platform. The per-platform breakdown — which Rankwize keeps separate rather than aggregating — determines which fix type to prioritize for which platform.

Finding 3: Most brands were invisible on the prompts that matter most commercially

Prompt Discovery scores intents on commercial value — how far into the buyer journey they sit. High-commercial-value prompts are things like "best [category] tool for [use case]," "alternatives to [competitor]," "[category] software comparison." These are the prompts where being cited converts to pipeline.

Across the 20 domains, commercial-value prompts had lower citation rates than informational prompts — consistently. The domains that ranked well organically on commercial keywords were not necessarily the same domains being cited on commercial AI prompts. The two rankings use different signals.

This is the core argument for AEO as a distinct discipline from SEO: ranking well in traditional search doesn't transfer automatically to AI citation, especially on high-intent queries.

Finding 4: The Update Existing gap has a short path to a fix

For the Update Existing cases — roughly 6 in 10 diagnoses — the recommended actions were consistently specific: add more structured evidence (stats, citations, case study data), strengthen the page's primary claim into a form an AI engine can quote directly, improve the metadata and H-tag structure to signal topical authority, and in most cases tighten the CTA to reduce ambiguity about what the page is for.

None of these require new content. They require editing. A content team running Rankwize against a domain with 6-in-10 Update Existing gaps doesn't need a new content calendar — it needs a sprint of page revisions.

What this means for your content strategy

Before you brief out new content to close AI visibility gaps, run diagnosis on what you have. The majority of the gap-closing leverage in this dataset came from pages that already existed, not from net-new content. New content creation is appropriate when the gap type is Create Comparison, Create How-To, Create Diagnostic, or when the matcher confirms no existing page is close enough to win — but that's a minority of cases.

The implication for how teams should sequence AI visibility work: audit your existing high-priority pages first. Identify which are failing on high-commercial-value prompts. Run diagnosis. Fix the fixable pages. Then build new content for the genuine gaps.

We'll publish more detailed breakdowns — with precise percentages and vertical-specific patterns — once we have customer data from live Rankwize tenants. The findings above reflect the baseline state before any AEO work; we expect the pattern to shift as we accumulate verified outcomes from the diagnose-fix-verify loop.