Self-assessment
Six checks you can run yourself in twenty minutes.
No score, no scan, no dashboard. These are the same first checks we run in a real assessment, with an honest explanation of what each result actually means.
Each check below takes two or three minutes and tells you something specific: what a stranger actually sees, what the search box suggests before they press enter, what three AI systems say when asked about you, and how much of your own first page you control.
The conclusion stays with you. If what you find is reassuring, you have spent twenty minutes and learned something. If it is not, you will have documented it yourself, in your own words, with dates, which is more useful than any score would have been.
The checks
In this order.
Search your name in a private window
How
Open a private or incognito window first, because your normal results are personalized by your own history and will look better than everyone else's. Search your full name, then your name plus your city, then your name plus your company.
What it means
Look at the first five results only. That is what almost everyone sees. If anything in those five would give a counterparty pause, it is materially affecting decisions already.
Watch the autocomplete before you press enter
How
Type your name slowly and read the suggestions the search box offers. Do it in a private window, and if you can, ask someone in another city to do the same.
What it means
A suggestion appears before a single result is read and frames everything after it. If a word appears there you would not choose, that is the single highest-cost item on this list.
Ask three AI assistants who you are
How
Ask more than one system, and ask each more than one way: "Who is [your name]?", "What should I know about [your name] before doing business with them?", "Has [your name] been involved in any controversy?"
What it means
Note whether the answers agree with each other. Contradiction between systems means your source material is inconsistent, which is fixable. Invented detail means the sources are too thin, which is also fixable.
Count how much of page one you control
How
For each of the first ten results, decide: is this something you published or can edit, something neutral, or something adverse?
What it means
Fewer than three controlled results on page one is a thin presence, and thin presences are where single negative items dominate and where models invent details.
Check whether someone else is being read as you
How
Search your name and look for anyone who shares it. Then ask an AI assistant about you and see whether details from their life appear in the answer.
What it means
Conflation is the most common cause of unfair machine descriptions we encounter, and structurally the most tractable. If you find it, it is worth addressing before anything else.
Look at what is dated
How
For each adverse item, find its publication date, and check whether the outcome, whether dismissal, settlement, resolution, or correction, appears anywhere as prominently.
What it means
An allegation that is visible while its resolution is not is a factual gap rather than an opinion problem, and factual gaps are the most correctable category there is.
Questions we are asked
- Why is there no automatic score?
- Because any number we generated would be invented. A single automated check cannot see personalized results, regional variation, or the distribution of AI answers across systems and phrasings, so a score built on one sample would be a guess wearing a decimal point. The six checks above are what we would actually look at.
- What if the results look fine?
- Then you are in a good position and the useful move is the proactive one: making sure it stays that way as the systems change. Most people who check are relieved, and checking annually is a sensible habit.
- What if the results are worse than I expected?
- Write down exactly what you found, with the queries you used and the date. That documentation is the beginning of a real baseline and it is genuinely useful whether or not you engage anyone.
- How is a real assessment different from this?
- It uses de-personalized, multi-region capture; runs a standard question set with several phrasings across systems; documents position and dated captures for every item; and produces a baseline that later change is measured against. This page is the version you can do yourself in twenty minutes.
Found something you did not expect?
Bring what you documented to a first conversation. It costs nothing, and we will tell you plainly whether it warrants action.