AI is most useful to a basketball coach when it starts from real data and ends with the coach's edit. Used that way, it can take seven jobs off your plate this season: a first read on the opponent's best scorer, matchup assignments, an attack plan against a coverage, teaching cues, a first-draft game plan, self-scout questions and plain-language explanations for players. It cannot know your roster, see effort, or make the decision for you. The coaches who get the most from it treat every answer as a first draft that saves an hour, not a game plan that replaces one.
| Job | What to ask | What you should get back | What you still check |
|---|---|---|---|
| Read their best scorer | Where does he score, and how do we take it away? | Shot profile, three defensive rules, where the ball goes on help | The sample behind each number |
| Matchup assignments | Who should guard their No. 1 option? | A primary defender, a backup, and why | Foul trouble and who is healthy |
| Attack a coverage | How do we attack a team that switches everything? | The advantage, first action, target, counter | Whether your team has practiced it |
| Teaching cues | Turn this tendency into three cues | Short phrases players can repeat | Your team's own vocabulary |
| First-draft game plan | Build a one-page plan from this scout | Keys, matchups, coverage, late-game calls | That it fits on one page |
| Self-scout | What would an opponent attack in our numbers? | Your weakest spots, in their eyes | Film of your last two games |
| Explain it to players | Explain drop coverage to a 15-year-old | A plain-language version with one example | That it sounds like you |
Start from data, not from a chatbot's memory
The difference between a useful AI answer and a dangerous one is where the numbers come from. A general-purpose language model asked about a player will happily produce statistics, and a 2025 study of language models tracking statistics through sports narratives found hallucination, omission and role confusion. The NFHS put it plainly in its own guidance on AI for school communication: humans must verify facts.
HoopBrief is built the other way around. Its answers are structured from public statistical and play-by-play data, and the free reports show their working. HoopBrief's How to Guard Anthony Edwards report, for example, shows that in 2025-26 he took 22.2% of his shots at the rim and finished 71.5% there, took 41.8% of his shots from three at 39.6% on 523 attempts, and got to the free-throw line about 7.2 times a game. The defensive rules that follow, wall off the paint early, chase over screens, do not reach on the drive, come straight from those numbers. That is what grounded AI looks like: every rule traceable to a number, and every number traceable to a sample.
What a good AI answer looks like
HoopBrief publishes a sample report on its free scout page, answering a question every coach faces in league play: "How do we attack a team that switches everything?" It comes back in five parts:
- The advantage. Switching trades matchups for mismatches. Find the worst defender and the slowest switch, and hunt them early in the clock.
- First action. An empty-side ball screen to force the switch, then attack before help loads from the weak side.
- Personnel target. Screen with the player guarded by their slowest-footed big, and re-screen if they switch back.
- Counter. If they pre-switch the screen, slip it. The screener's defender is now out of position at the rim.
- Walkthrough focus. Late-clock spacing and the read on second-side help.
Notice the shape. It starts with the point, gives one action rather than ten, names a target, plans for the counter and ends with what to practice. That is the shape FIBA's coaching manual asks for when it tells coaches to give players three or four key things and a cure rather than a diagnosis. Hold any AI answer to that standard, and try it on a matchup of your own with the free HoopBrief scout.
The seven jobs, one at a time
1. Read their best scorer
Ask where he scores and how to take it away. You want a shot profile by zone, three defensive rules and where the ball goes when you help. That is a 30-minute film job reduced to a two-minute read you then verify on two clips.
2. Matchup assignments
Ask who should guard their first option and why. A good answer weighs size, foot speed and foul risk, and gives you a backup. Your job is to add what the data cannot know: who is healthy and who is in foul trouble.
3. Attack a coverage
Coverages are where AI saves the most thinking time, because the answers follow patterns. Ask how to attack drop, switch, ice or zone, and you get a first action and a counter, as in the example above.
4. Teaching cues
Paste in a tendency and ask for three cues a player can repeat in a timeout. "No middle", "top-lock the shooter", "stunt and recover" beat a paragraph every time.
5. A first-draft game plan
Ask the tool to turn your scout into a one-page plan, then edit it against our game plan template: three keys, matchups, one coverage, their top two actions, your offense, a glass rule and late-game calls.
6. Self-scout
Ask what an opponent would attack in your own numbers. It is uncomfortable and it is the most valuable question on this list, because it shows you the scouting report the other bench is writing about you.
7. Explain it to players
Ask for a plain-language explanation of a concept for a 15-year-old, with one example. Then rewrite it in your own words, so it sounds like you, not a manual.
The three jobs AI cannot do
Know your roster. It does not know who rolled an ankle on Tuesday, who is in foul trouble by the second quarter, or which set your point guard runs best.
See effort. Body language, communication on defense and how a player responds to a bad call are visible on film and in practice, not in play-by-play data.
Decide. AI can give you three good options. Choosing one, and committing the team to it, is coaching.
Guardrails that keep AI honest
- Trace every number. If a claim changes a coverage or a matchup, find the sample behind it. Our guide to validating an AI scouting report gives a five-check method.
- Label sample sizes. A tendency seen in five games is a plan. One seen in one game is a question for film.
- Protect player data. Do not paste students' personal information into general-purpose tools. The NCAA's performance-technology guidance, approved in December 2025, asks college programs to have a written plan for data protection when they adopt software and apps, and high school programs should hold themselves to the same standard.
- Keep a coach between the tool and the players. Nothing reaches a player until a coach has edited it.
Put it to work this week
HoopBrief is AI basketball analysis built for exactly these seven jobs. Ask a basketball question and get a structured, scouting-style answer from public statistical and play-by-play data, read through the coaching lens you choose. It does not host or tag film, and it does not replace your eyes. It gives you the first draft in seconds so your time goes into the decision.
| Starter, $9.99 a month | Pro, $99 a month | |
|---|---|---|
| Questions | Unlimited | Unlimited |
| Matchup reports | Core report with attack and guard plans | Deep report |
| Coaching lenses | 5, including Matchup Hunter and Analytics | All 12 |
| Positioning IQ cues and micro-behavior reports | Not included | Included |
| Saved reports | 10 | Unlimited |
Coaches who prepare every week usually choose Pro for all 12 lenses and unlimited saved reports; Starter is a strong way to start. Both are billed monthly and cancel in one click. See plans and get started, or run the free scout on your next opponent's best player first. For ready-made questions, see our scouting prompts before your next game.