Why a 100 is not AI ready: relative vs absolute scoring

Two dealers in the same market open two reports. One says the store is a 100. The other says it is a 41. Both are accurate. They are measuring different things, and only one of them tells the dealer what to do next.

Relative scoring

A relative score ranks stores against each other. The best store in the lens is set to 100 and everyone else is scaled to it. It is easy to read and it makes a good leaderboard, which is why public brand-by-metro reports use it.

The problem is that it cannot answer "am I ready." If every Mercedes store in Atlanta is missing an inventory feed and an MCP endpoint, the relative leader is still a 100. The number says "you are winning." An AI agent trying to check inventory says "nothing here."

Absolute scoring

An absolute score measures against a fixed standard. On our readiness scale, 100 means every check passed: schema, crawler access, llms.txt, AI discovery, machine-readable inventory, MCP endpoint, sitemap, entity match. A store scores what it has, regardless of who is next door.

Across 2,699 scored franchise rooftops in September 2026, the average is 40.7 and the best is 62. That is the honest state of the industry. A relative scale would have handed out a few hundred 100s.

Why dealers get hurt by relative scores

It stalls investment. A GM who sees a 100 sees no reason to spend. Six months later a competitor deploys the agent layer, and the GM's 100 becomes a 60 without anything changing on his site.

It hides the fix. A relative score is a rank. It cannot be decomposed into "you are missing X." An absolute score is a checklist with points attached.

It rewards the wrong thing. Stores climb a relative leaderboard when competitors get worse. Stores climb an absolute scale when they get better.

When relative scores are useful

For visibility, not readiness. "What share of AI answers name my store versus the competitor" is a relative question and a fair one. We report share of voice and named first exactly that way. The mistake is applying a curve to the readiness side, where the standard is fixed and the ceiling is real.

How to read any AI report

Ask three questions. Is the score relative or absolute? What would a 100 require? Can I see the raw answers or the checks? If the vendor cannot answer all three, you are looking at a leaderboard, not a measurement.

What is a good absolute readiness score today?

Above 60 puts a store in the top handful of thousands we have scanned. Above 80 requires the agent layer, which almost nobody has yet.

Does PromptGraph ever use relative scores?

Yes, for visibility metrics like share of voice and named first, where the question is inherently comparative. Never for readiness.