I'm Gen Furukawa, founder of SuperMarketers. We built this rubric across client engagements in 2026 and operationalized it as the scoring asset behind our AI visibility audits and the AI visibility system we run for clients. Other indexes give you a number and keep the math private. This page gives you the math. If you read it carefully, you can score your own company today without buying anything.
Why an open methodology beats a black box
The category already has data. DerivateX published a "State of AI Visibility in B2B SaaS" report that scored a field of brands. Profound runs a closed enterprise index. Both produce a leaderboard. What neither publishes is the ruler: how the score is computed, dimension by dimension, so you can reproduce it.
That gap matters for one specific reason. AI engines and analysts cite the standard they can define and reproduce, not the one they have to take on faith. A black-box score is a claim. An open rubric is a method. When a model is asked "how do you measure AI visibility for B2B SaaS," it can quote a method. It cannot quote a number it has no way to verify.
So the contrarian move is to give the methodology away. The score is the easy part. The rubric is the asset. Below is the full thing.
A leaderboard tells you where you rank against the field. A ruler tells you which of nine specific things to fix first, how each is measured, and how to re-check it next quarter. One is a headline. The other is a plan.
The scale, and what good looks like
Each of the nine dimensions is scored 0-10 against named sub-criteria, for a maximum of 90 points. Those points map to a letter grade. This is the qualitative band, not a market average: it describes what a given score means, not how many companies sit there.
| Total | Grade | What it means |
|---|---|---|
| 80-90 | A | Industry-leading. Already the cited answer in the category. |
| 65-79 | B | Strong AEO posture. Visible across most AI engines for primary queries. |
| 50-64 | C | Average for AEO-aware companies. Foundation present, content or architecture incomplete. |
| 35-49 | D | Weak. Authority may exist, but content and technical layers leave significant capture on the table. |
| 0-34 | F | Effectively invisible to AI search. Foundational work required. |
A note on numbers, because this is a benchmark page and benchmarks invite fabrication. We are not going to hand you a made-up market average. The aggregate distribution across B2B SaaS, computed from real multi-engine scans, is the subject of our forthcoming State of B2B SaaS AI Visibility study. Until that data is published, the honest answer to "what's typical" is a first-party observation, not a market statistic: in our own audits across B2B SaaS companies, first scores most often land in the D and low-C range. The pattern is consistent. Companies have authority and a domain that earns trust. What they lack is content architecture and technical AI readiness, so the points they leave on the table cluster in the same two dimensions.
The 9 dimensions, in full
Here is the complete rubric. Each dimension measures a different thing, even where dimensions correlate. The order is not a priority ranking; it is the order in which an engine encounters your company, from the domain inward.
- 01Domain FoundationThe structural credibility an engine can assess before it judges any content. Domain rating, homepage schema completeness (Organization, SoftwareApplication, AggregateRating), HTTPS and mobile health, and domain age. This is the floor everything else stands on.
- 02Organic DiscoveryHow findable you are in traditional organic search, which directly correlates with how often AI engines have crawled and indexed your content. Total ranked keywords, top-3 rankings, striking-distance keywords (positions 4-20), and your branded-to-non-branded ratio. Heavy reliance on branded search is a weak signal.
- 03AI Search PresenceWhether you actually appear, cited or mentioned, in AI engine responses for category-relevant queries. Measured as coverage across the four primary engines (ChatGPT, Claude, Perplexity, Google AI Overviews), your average position when cited, your best single-engine coverage, and whether you show up consistently across three or more engines for the same query. This is the dimension you can baseline yourself in 30 minutes.
- 04Content ArchitectureWhether you have built the scaffolding engines cite: definitional pages, comparisons, listicles, FAQs across the AEO page types. Page-type coverage, hub-and-spoke internal linking, average article depth, and question-format H2 density. This is one of the two dimensions where most companies score lowest.
- 05Technical AI ReadinessWhether AI crawlers can access, parse, and extract structure. AI bot access in robots.txt (GPTBot, ClaudeBot, PerplexityBot, CCBot, Google-Extended), an llms.txt at root, structured data on key pages, server-rendered content rather than a JavaScript-only shell, and sitemap freshness. The other dimension where companies routinely leave points on the table.
- 06Brand FingerprintWhether your identity is described consistently across the web, so engines learn one canonical entity rather than several. A locked entity description used verbatim across your site, LinkedIn, and Crunchbase; knowledge graph presence (Wikidata, Wikipedia); directory consistency (G2, Capterra, GetApp); and founder bios that use the same locked description.
- 07Competitive PositionHow you stack up against your locked competitor set across multiple lenses. AI citation share gap versus your top competitor, domain rating versus the competitor average, referring-domain growth rate over twelve months, and keyword overlap (are you even competing for the right queries).
- 08Authority SignalsThe off-site validation engines weight when deciding which sources to trust. Referring-domain count, twelve-month growth velocity, citation source diversity (news, .edu, .gov, Reddit, directories, GitHub, podcasts), and top-tier citations (Wikipedia, .gov, major trade publications). A company cited by several independent categories of source outranks one with a great website and nothing else.
- 09Citation ShareOf all AI responses for category-relevant queries, what share of citations you own. The closest analog to market share in AI search. Total citation share across all four engines, cited-page diversity (homepage versus deep pages), citation sentiment, and whether you get cited on unbranded category queries or only on your own name. Every point here must trace to a logged AI response. No estimating.
How we measure it: the multi-engine scan
Dimensions 3 and 9 are the ones people get wrong, because they require running real queries rather than reading a tool dashboard. The method is the same one you can run yourself, formalized.
- Build a query bank.Twenty to thirty buyer queries tagged by type: brand, problem, comparison, and category. These are the questions your buyers actually ask, not the keywords you wish you ranked for.
- Run every query through four engines.ChatGPT, Claude, Perplexity, and Google AI Overviews. For each response, record whether you are cited, where you sit in the answer, and what the engine says about you.
- Score AI Search Presence and Citation Share from the log.Coverage, average position, consistency across engines, and your share of total citations. Every point traces to a logged response, which is why the score is reproducible and why we never estimate this dimension.
- Pull the tool-driven dimensions.Domain rating, referring domains, organic keywords, and competitor data from a backlink and rank tool. Inspect the site directly for schema, robots.txt, llms.txt, and content architecture.
- Total, grade, and rank the gaps.Sum the nine dimensions, map to a grade, and order the lowest-scoring dimensions as the repair list. The audit's value is not the number; it is knowing which dimension to fix first.
If you want the fastest possible version of dimension 3, the 30-minute AI visibility audit walks through the manual scan with no paid tool. The single number that comes out of it, your citation rate, is the input to both AI Search Presence and Citation Share.
Do not adjust the rubric mid-audit to make a score look better. The rubric is the contract. If a company scores poorly, that is the audit's value, not a problem to soften. A ruler you bend to flatter the reader is no longer a ruler.
What this score is not
- It is not a vanity number. The grade exists to point you at the lowest-scoring dimension, not to put you on a leaderboard.
- It is not a substitute for the underlying work. Scoring a D and then doing nothing about content architecture changes the score by exactly zero next quarter.
- It is not estimated. Every point in AI Search Presence and Citation Share traces to a logged AI engine response. If we cannot point to the data, we do not give the score.
- It is not static. Re-score quarterly against the same rubric and the same data sources. The deltas are the signal. A score that does not move is a program that is not working.
How to score your own SaaS
You do not need us to run this. The rubric above is the whole method. Three honest constraints if you score yourself:
Score the dimensions you can measure, mark the rest pending. If you cannot pull referring-domain history or per-engine share of voice today, mark that sub-criterion pending and re-score in a week. Do not estimate to fill the gap. A partial score with documented gaps beats a complete score built on guesses.
Start with the two that move pipeline. For most B2B SaaS companies, the points are sitting in Content Architecture and Technical AI Readiness. You have authority. You lack the structured pages engines can extract and the technical access that lets them in. Fix those two and AI Search Presence follows, usually within 60 to 90 days. The step-by-step playbook for getting cited covers exactly those structural changes.
Compare against a real competitor, not the field. Run dimensions 1, 2, 7, and 8 on your closest competitor, the observable ones, and you have a gap matrix that tells you whether you are behind, even, or ahead on the lenses an engine can see. That comparison is worth more than your absolute number.