MachineRead terminal-window signal markA stylized terminal window with a green signal scan-line and two data marks, indicating machine-readable signal inspection.MachineReadAI & search readiness audit

Blog

What a MachineRead benchmark compares

Read benchmark bars with their preset, denominator, and snapshot date so a peer comparison answers the question you intended to ask.

author
published
reading
4 min

A benchmark helps put an audit score in context, provided the comparison fits the site. A blog and an online store may expose different optional surfaces; placing them on the same chart does not make their requirements identical.

MachineRead compares your result with stored peer audit snapshots. Read the scope, score denominator, and snapshot date alongside the bars. The chart concerns evidence from the audit, not the peers' traffic, revenue, or search rankings.

Same preset means same stored peer group

An unmodified named preset uses its corresponding stored peer group. A Blog/Content audit therefore has a more specific comparison than a Custom audit that mixes families from several site categories.

Custom, legacy, and materially changed named scopes use disclosed broader context rather than claiming a like-for-like cohort. That distinction should remain visible when sharing a chart. A customized selection of a few API checks is not necessarily equivalent to the full SaaS preset.

The peer evidence is stored rather than fetched again during your audit request. The snapshot date identifies the comparison's age; it does not establish what every peer returns today.

Percentiles locate a report inside the returned snapshot

The Essentials benchmark uses the included, non-locked rows. Its percentage is distinct from raw points on the full rubric: an illustrative 42 earned points out of 60 checked points produces 70%, not 42% (MachineRead, "API Reference").

The median and percentile then place that percentage within the returned peer pool. They describe that pool, not a representative sample of every website. Read the snapshot date before comparing an earlier result with a later one: the target may have changed, the stored peer evidence may have changed, or both.

Use the comparison to decide which findings to inspect. It does not establish that a higher-scoring peer has better search rankings or receives more agent traffic.

Strict agent readiness is a separate lens

MachineRead also has a strict agent-readiness comparison. It should not be blended into the 60-point Essentials evidence score.

The strict readiness path counts passed labels that are present in the current scope's probe list. Its maximum starts with a base set and adds protocol, account/auth, and ecommerce probe families when those families are included. The result has its own percentile and position labels, based on strict-readiness percentages rather than the Essentials evidence score.

That makes strict readiness a separate discovery-signal lens. It can say how a report compares on explicit agent-discovery probes under the resolved family scope. It cannot say that an agent will adopt the site, complete a task, cite it, rank it, or send value back to the publisher.

A practical example

Consider two fictional reports.

Report A uses the unmodified Ecommerce preset. The backend returns ecommerce-group profiles, so the default comparison is same-preset peer context. If the report shows a percentile, that percentile is relative to the returned ecommerce stored-audit snapshot and its denominator.

Report B starts from a named preset but changes enough scope families that the returned context is multi-group. In that case the client should default to disclosed all-profile context, not pretend the pool is an Ecommerce-only cohort. The number can still orient the reader, but it answers a broader-context question.

What to check when reading the benchmark

Begin with the selected score: Essentials evidence or strict agent readiness. Then check the cohort label and snapshot date. A nearest peer is a point of reference in that returned pool, not necessarily a business competitor.

When repeating an audit after a site change, compare the underlying rows as well as the relative position. A useful improvement is one you can explain, such as fixing a contradictory canonical or restoring a readable discovery document. Moving on the chart without understanding the changed evidence is much less informative.

Works Cited

MachineRead. "API Reference." MachineRead, https://www.machineread.ai/docs/api. Accessed 10 Sept. 2026.

See also