The short version
Coming soon: this survey is built and will appear in the app once it's published.
People who use Claude at work rate themselves in 10 skill areas, then answer three short quiz questions in each. Every question has one right answer, so the level comes from what people got right, not from how they rate themselves. The whole thing takes about 20 minutes.
Each person gets a level from Beginner to Expert, and the report shows gaps by skill area and by team, and where confidence and results differ. It is not an Anthropic certification.
Ten skill areas
The questions ask about ways of working and long-standing capabilities, never about a model, a price or a menu, so they stay valid as Claude changes. Each area opens with the same self-rating question, then its three quiz questions.
Before the areas, About you asks for a team (optional, from a list set per round), "How often do you use Claude?" (Less than monthly, Monthly, Weekly, Most working days) and "How long have you used Claude at work?" (Under 3 months, 3–11 months, 1–2 years, Over 2 years), so the report can set use and experience against the score.
A quiz, not a self-rating
Self-ratings measure confidence. They are useful, but a team can be confident and wrong, or competent and modest. So the level comes from a quiz: three questions per area, one each at a basic, practical and advanced level, each with four answers, one of them right, plus "I’m not sure". A right answer is 1 point; a wrong one, or "I’m not sure", is 0.
The real questions aren't published, so the quiz stays fair from one round to the next. Here is one of ours in the same shape, with its answer:
- The self-rating comes first, so the quiz questions can't change how people rate themselves.
- No time limit, and people can go back and change an answer before they submit. It isn't supervised, so read it as a skills check, not an exam.
- The right answers stay on the server. They never reach the survey, and people never see them, not even in their own result. Whoever can open the report sees them, ticked, on the Questions page.
- Every question was checked against Anthropic's documentation on 9 October 2026. When a right answer has to change, the question gets a new identity, so rounds compare only questions that stayed the same.
From right answers to a level
The score is right answers out of 30, and the level comes from it. The same cut-offs apply to everyone, whatever they rated themselves.
Move the score to see the level it gives:
An area is scored on its own too, as right answers out of 3. The claude code and the api questions (below) are scored separately and never count toward the level.
Confidence against results
The self-rating isn't scored. It's set against the quiz, area by area, to show where confidence and results part ways:
In the report's Confidence and quiz section, each area shows how people rated themselves next to the share of answers they got right. A person's result also groups their areas as Strong (all answers right), Getting there (some) and Start here (none yet), and tags their two strongest areas.
Extra questions
One question in About you decides whether someone gets extra questions at the end:
Several answers are allowed, so someone who uses both gets both sets. The survey hides questions that don't apply, and the server refuses answers to them.
What people see
Whoever launches the survey decides with one switch, Show people their own results, on by default: "Right after they submit: their level and a score per area. Never the right answers." Off, people see a thank-you message only.
- Their own result, right after submitting: the level with its one-line summary, each area as right answers out of 3, and where their self-rating and the quiz disagree. With it a private link: "Save this link to see your result later. Anyone with it can see it." It keeps working after the survey closes, and stops if results are switched off.
- What they never see: which questions they got wrong or the right answers, the claude code and the api scores, other people's results, or the company's numbers. The result says so: "Your company can see this result. The right answers aren't shown, so the questions stay fair for the next round."
What the report shows
The report fills in from the first response and has three pages: Overview, Questions and People.
- Top summary: the average score out of 30 with the share of answers right, how many people are at Practitioner or above, the biggest gap (the area with the lowest share of right answers), and how many people took it.
- Overview: how many people reached each level; the share of right answers in each area; a heat map of teams by area, from "Under 35%" to "85% or more" right; Confidence and quiz; Use and skill, "The average score of people who use Claude this often.", and by what they use claude through; and the Claude Code and the API questions on their own.
- Questions: every question with the share that picked each answer, the right one ticked, and the most common mix-up.
- People: everyone who answered, sorted by name, never ranked, each with their score, areas and blind spots. Open a person for the result they saw.
Trends appear after the second round: the average score and the share at Practitioner or above, round by round, for the same version of the questions.
Who sees what
- It can run anonymously. The one-click launch uses a shared password with no names. Advanced setup can ask for a name or an email, or use Google sign-in, and then the report names people, with the People page and Responses open to whoever can open the report.
- Nobody is ranked. People are sorted by name. The public link shows the Top summary and Overview only, and the AI connection returns aggregates, never answers or names.
- For learning, not for decisions about people. As the survey's own guidance says: "Results are for learning and planning, not for hiring or firing decisions on their own."
Limits
- It is not an Anthropic certification, and Anthropic hasn't made or endorsed it. It isn't supervised, and three questions per area is a quick check, not a full examination: it tells you where a team stands and where to start, not whether one person is qualified.
- Small groups swing. Results show from the first response, with no minimum group size, so a small team's numbers can point to one person.
- The questions date. They ask about ways of working rather than menus and prices, and were checked against Anthropic's documentation on 9 October 2026; we re-check them and replace what has changed, which starts a new trend line for those questions.