Decide what the evidence is enough to change.

Before adding a task to the plan, name the buyer problem it should solve. Then ask whether the evidence points to that problem, how costly the change is and what you would check afterwards.

Act, investigate or leave it alone

The following are decision examples, not findings about individual businesses in the June sample.

Situation	A proportionate next move
A current check finds that a main enquiry link is broken	Repair the destination, test the completed action and confirm receipt.
Buyers repeatedly ask what a service includes	Check whether the page answers that question clearly; improve the explanation where it is missing.
A public check suggests an important page is restricted	Confirm the intended setting and inspect current access/indexing evidence before changing controls.
A service page has no table or download	Leave it alone unless that format would help a real buying decision.
A page has a low DUCR total but no identified buyer problem	Inspect the category and underlying observations before assigning work.

For page structure, use the service-page guide. For evidence selection, start with the concern your proof needs to answer.

Match the question to the evidence

Question	Evidence needed beyond this study
Are platforms actually citing us?	Recorded questions, answers, dates and cited URLs, with the sampling conditions kept visible
Are relevant visitors reaching our pages?	Search and analytics records interpreted against the intended audience
Are enquiries arriving and fitting the business?	Confirmed receipt, enquiry details and business follow-up records
Did a particular change cause an improvement?	A comparison that accounts for timing, other changes and alternative explanations
Should we rebuild the site?	The specific structural or technical obstacle, repair options, cost and business priority

The monthly answer log supports a defined sample of observations. The business scorecard connects visibility to received enquiries. Neither makes a single observation a platform-wide result.

Read the June study in its actual context

Study characteristic	How to use it
50 selected records	Describe this group; do not turn it into a market-wide prevalence estimate.
Accessible public sites were selected	High access counts cannot tell you how many inaccessible sites exist elsewhere.
35 consulting, accounting and agency records	The mix is uneven, with all original country fields coded US; it does not establish which industry or market performs best.
Public-page observations	No private Search Console, analytics, server logs, CRM or sales outcomes were measured.
28 records retained with caveats	Keep the review labels when calculating or quoting findings.
Direct identifiers removed; source pages not supplied	You can check record arithmetic. The corrected export does not provide source pages; its stable IDs may still be linked to earlier downloaded versions.
June collection dates	Recheck current conditions before applying a historical observation to a live website.
Our own scoring weights	Treat the score as a review framework, not a calibrated citation or conversion prediction.

The study contains no actual AI citation tracking. Its access checks and content flags cannot establish why an assistant selected or omitted a business.

Turn a finding into a useful question

Recorded finding: methodology was marked present on 10 of 50 reviewed pages.

Question for your business: when a buyer wants to understand how the work will be done, can they find a clear explanation?

Possible next move: improve the relevant process explanation if it is missing and matters to that decision. If the answer is already clear, a section with a particular heading is not a reason to add work.

That is the link we look for between research and a recommendation: an observation, a business concern and a check that can confirm whether action is useful.

Keep a finding attached to its edition

The June 2026 records and rubric remain the study’s historical edition. The October 5 update corrects interpretation and the public export’s identifying fields while retaining observations and scores. The dataset record explains the correction.

A later sample, scoring revision or collection should be identified separately. New platform guidance should carry its own review date; it should not be presented as something measured earlier.

Bring us the decision, not just the score

For a public review of your website, request the free Outside View. If you already know the question you need to resolve, request a Clarity Call. We can discuss whether the next move is a focused repair, clearer information, further investigation or no immediate change.