How to Measure AI Visibility for B2B.

Most B2B companies have no idea whether AI recommends them. Measure AI visibility before you optimise for it, because without a baseline you cannot tell whether your work is moving the needle or burning budget on assumptions.

Traditional SEO gives you rank positions and click-through rates. AI visibility is different: there are no SERPs, no rank numbers, and no click data. The output is a paragraph that either names your brand or does not. That binary outcome, multiplied across hundreds of buyer prompts and four major engines, is what you need to track.

Why does AI visibility matter for B2B companies?

AI visibility matters because your buyers are already using ChatGPT, Perplexity, Gemini, and Microsoft Copilot to shortlist vendors before they ever visit your website. If you are not cited in those answers, you are invisible during the research phase that shapes the shortlist.

The shortlist is real: in SourceRank AI's October 2026 study of 22 B2B categories, 98.6% of 1,249 AI answers named at least one firm, and the most named firm in each category appeared in 21.8% to 93.0% of answers (who AI recommends). The companies at the top of those lists are capturing the early mindshare that turns into pipeline.

What are the four engines you need to track?

Not all AI engines behave the same way. Each has different source preferences, citation patterns, and update cadences.

Understanding these differences is critical. A company that ranks well in Perplexity (which favours fresh, well-structured web content) may be invisible in ChatGPT (which leans on training data and established authority). Your measurement approach needs to cover all four.

How do you build a prompt library for measurement?

The foundation of AI visibility measurement is the prompt library: a structured set of questions that represent how your buyers actually research solutions. Building it requires three steps.

1. 1. Map the buyer journey from problem awareness through vendor evaluation to final decision
2. 2. Write 50-100 prompts that mirror real buyer questions at each stage, varying phrasing and specificity
3. 3. Classify each prompt by intent (informational, comparative, transactional) and by persona (technical evaluator, executive sponsor, procurement)

The prompt library should be refreshed quarterly. Buyer language shifts, new competitors enter, and AI engines update their behaviour. A stale prompt library gives you stale data.

What metrics should you track?

Forget vanity metrics. AI visibility measurement comes down to three numbers that tell you whether your brand is being recommended.

1. 1. Share of Voice (SoV): the percentage of prompts where your brand is cited, out of the total prompt library
2. 2. Citation quality: whether you are named as a primary recommendation, a secondary mention, or an "also consider"
3. 3. Engine coverage: your SoV broken down by engine, because optimising for one does not guarantee visibility in the others

Track these monthly. Plot the trend. The absolute number matters less than the direction: are you gaining or losing share as competitors invest in the same space?

How do you run the measurement process?

Running AI visibility measurement manually is possible but tedious. Here is the process, whether you do it by hand or have SourceRank AI run it for you.

1. 1. Submit each prompt in your library to all four engines
2. 2. Record whether your brand is mentioned, in what context, and what competitors appear
3. 3. Score each response: cited (primary), cited (secondary), not cited
4. 4. Aggregate into SoV by engine, by prompt category, and overall
5. 5. Compare against your previous measurement period
6. 6. Identify the gaps: which prompt categories and which engines are you weakest in

The gap analysis is where the value lives. It tells you exactly where to invest: if you are strong in Perplexity but invisible in ChatGPT, the fix is different than if you are weak everywhere.

What are the most common gaps B2B companies find?

Across the B2B brands we have audited, certain patterns repeat.

• No authoritative third-party mentions: the brand exists only on its own website, so AI engines have no external signal to trust
• Thin entity presence: no Wikipedia page, no Crunchbase profile, no industry directory listings
• Content answers questions AI never sees: blog posts exist but are gated, JavaScript-rendered, or blocked by robots.txt
• Competitor dominance: one or two competitors appear in a large share of answers, leaving little room; in the most concentrated category in SourceRank AI's October 2026 study, small business insurance, the three most named firms took 46.1% of all mentions

Each gap has a specific fix. Third-party mentions require off-site authority building. Entity presence requires structured data and directory listings. Content accessibility is a technical fix. Competitor dominance requires a sustained investment in both content and authority.

How long does it take to improve AI visibility?

AI visibility moves slower than paid search but faster than traditional SEO. How fast it moves depends on the gap: indexing and schema fixes need a recrawl, while off-site corroboration takes longer to build. Judge progress over several measurement runs rather than one, because the sources engines cite move between runs even when nothing on the web has changed (what AI engines cite).

The timeline depends on your starting position. A company with strong existing content and some third-party mentions will move faster than one starting from zero. The pricing page outlines what a typical engagement looks like.

Frequently asked questions