CEO Reputation Audit Scorecard: 5 Engines, 1 Page for 2027

A CEO reputation audit scorecard turns the answers from five AI engines into one page: a score for accuracy, completeness, tone, sources and peers, a change since last month and a ranked list of fixes. Muck Rack's December 2025 report warns that citation behavior changes as models update, so a repeatable scorecard matters more than a one-time look.

By Ronn Torossian. Last updated: November 11, 2026. The scorecard is my own template, not an industry standard. The citation points are attributed to Muck Rack's December 2025 and May 2026 reports, and the scored example is a hypothetical for a fictional CEO.

What does the one-page scorecard contain?

The scorecard contains one row per engine and five scores per row. The table is the layout, and each cell takes a score of 0, 1 or 2.

How is each score decided?

Each score uses a simple rule so that two people would score the same answer alike.

• Accuracy: 2 if every fact is correct, 1 if one is wrong, 0 if two or more are wrong or a claim is invented.
• Completeness: 2 if the main achievements appear, 1 if some are missing, 0 if the answer is mostly empty.
• Tone: 2 if the description is fair, 1 if it is neutral or dated, 0 if it is unfair or negative without basis.
• Sources: 2 if cited pages are credible and current, 1 if mixed, 0 if outdated or low quality.
• Peers: 2 if the engine places you in the right category, 1 if partly, 0 if wrongly.

What does a scored scorecard look like?

A scored scorecard shows each engine's five scores and its total. The example below is a hypothetical for a fictional CEO and is not a real result.

The five totals add to 36 out of 50, an average of 7.2 per engine. The weakest row, Google AI Overviews, scored 0 for accuracy and 0 for sources, which points straight to the page to fix: the engine is relying on an outdated or low-quality source that states a wrong fact.

What goes under the scores?

Under the scores goes a short list of fixes ranked by risk. Invented claims about conduct come first, outdated facts second and missing achievements third, as described in the guide to what AI gets wrong about leaders. For each fix, name the source page, the owner and a date.

For an outdated fact, the guide to outdated AI information explains how to trace and correct the page. The first audit is the baseline, and the CEO AI reputation audit lists the five questions to ask.

The ranked list in the table is a hypothetical template. Every row names a page, an owner and a date, which turns a score into work.

How should the scorecard be used each month?

The scorecard should be repeated monthly with the same questions and compared with the previous page. A rising total shows that fixes are working, and a falling total points to a changed source or a new error. Keep each month's page in a shared folder, and brief the CEO on one line: the total, the biggest change and the top fix.

Muck Rack's December 2025 report warns that citation behavior changes as models are updated, so a drop does not always mean something you did. The AI brand monitoring guide and the Citation Share definition describe tracking over time.

What does a three-month trend look like?

A trend table shows the total across months and the change from the previous month. The numbers below are a hypothetical continuation of the example above.

The third month fell, and the table gives a reason. That is why the log records why a score moved and not only the score. Without a reason, a drop looks like a failure of the work when it may reflect a change in the engine.

How do you keep the scoring honest?

You keep the scoring honest by running each question twice and having a second person score a sample. Muck Rack notes that asking the same question again can return a different set of citations, according to Unite.AI, so one run can overstate or understate a score.

1. Run each question twice in each engine and score the lower of the two.
2. Have a second person score one engine's answers without seeing the first score.
3. Compare the two sets and settle differences by the written rules above.
4. Keep the saved answers, so a later reader can check a score.

What can the scorecard not tell you?

The scorecard cannot tell you how many people saw an answer or whether it changed anyone's mind. It measures what the engines say, not what readers do. Pew Research Center's March 2025 browsing data from 900 U.S. adults found that users clicked a traditional result in 8% of visits when a Google AI summary appeared, so many readers see only the summary, but no study measures the effect of a wrong summary on a specific person.

Treat the score as a quality measure of the record engines can read. It is a guide to which pages to fix and a way to see progress, not a measure of reputation.

Frequently asked questions

What is a CEO reputation audit scorecard?

A one-page grid that scores what five AI engines say about a CEO on accuracy, completeness, tone, sources and peers.

How often should the scorecard be updated?

Monthly, using the same questions, plus an extra check before a fundraise, interview or announcement.

Who should own it?

The head of communications, with a named backup, reporting one line to the CEO.

What is a good total score?

There is no standard. As a rule of thumb, a total of 9 or 10 on an engine means little to fix, and a total of 5 or less means the answer needs work. Judge by the trend in your own scores, not by a fixed target.

Should the CEO see every engine's answer?

The CEO should see the one-page summary and any answer that scored 0 on accuracy. The full set of answers should stay in the shared folder for the team and for counsel.

Copy the table into a spreadsheet and score today's answers. That first page is your baseline.