Aquatic directors evaluating AI drowning detection vendors should ask five things above all: does the system alert on early distress or only on a completed submersion; will it run on the cameras already installed or does it demand proprietary hardware; how long does a site take to go live; how is footage stored, retained and access-controlled under GDPR and the UK Data Protection Act; and how does an alert physically reach a lifeguard standing on poolside. Those five questions separate a genuine AI pool safety system from a camera with a motion trigger. This guide is written for aquatic operations and safety leaders at multi-site commercial pool operators — leisure and fitness chains, local-authority and leisure-trust estates, community-pool networks, universities and hotel groups — the people accountable for both the lifeguard roster and the risk that sits under it.
The questions matter because the answers vary enormously between suppliers. Some systems require dedicated underwater cameras and a construction programme; Lynxight is camera agnostic, connecting to standard overhead security cameras across roughly 10-12 manufacturers and models and covering every tile of the water from at least two angles. Some take a full swim season to commission; Lynxight states it brings a site live in about 50 days on average, and as fast as 2-3 weeks, against the 3-5 months typical of competitors that require dedicated hardware. And some vendors are vague about data handling, while Imperial College London publishes a public description of its Lynxight installation at the Ethos swimming pool, including the policy that footage is automatically deleted after 7 days unless it is needed for incident review — exactly the kind of documented answer a procurement panel in 2026 should expect to see before signing.
Throughout, one framing should guide the conversation: any credible supplier in this category is a decision support system, meaning it supports the lifeguard's judgement rather than acting on its own. It never enters the water, and the lifeguard remains the responder. The sections below turn that principle into a practical question set covering detection philosophy, integration, data protection, alerting, operational intelligence and evidence you can verify.
Which questions reveal an AI drowning detection vendor's real detection performance?
Narrow the questions to detection performance alone — not dashboards, not pricing — and a short list of technical asks will reveal how an AI supervision vendor actually behaves on the water. Ask for definitions before you ask for numbers, because the same claim means different things depending on what the vendor counts as an event.
What terms should the vendor define on the record?
- Computer vision — software that interprets live camera video rather than simply recording it. Ask which video streams it uses, at what resolution, and whether it needs proprietary underwater hardware.
- Pose estimation — the technique of mapping a swimmer's body points (head, shoulders, limbs) to classify posture and movement. Ask whether classification depends on posture alone or also on trajectory over time.
- Distress event — the behaviour that triggers an alert. Ask for the full alert taxonomy: does it include the instinctive drowning response, the involuntary early-stage behaviours that precede submersion, or only a body already motionless on the pool floor?
Which performance attributes should be specified?
| Attribute | What to ask for | Why it matters |
|---|---|---|
| Alert trigger point | The earliest behaviour that fires an alert | Prevention over detection — early notification before an event escalates — is a different product from post-submersion detection |
| Time-to-alert | Measured from which observable moment | An alert after full submersion buys your team far less time |
| Tested scenarios | Face-down and face-up submersion, breath-holding, shallow water blackout, glare, surface chop, crowded lanes | Real pools are busy, bright and reflective |
| Water coverage | Angles per area of water, and blind-spot handling | Uncovered water is undetected water |
| Reliability profile | Typical alert volume per pool per day | Sets realistic expectations for guards |
Then ask for operators at your scale. More than 50 BlueFit pools run Lynxight as standard, according to BlueFit CEO Todd McHardy — reference sites that carry your own daily conditions are stronger evidence than any lab demonstration.
How can aquatic directors verify alert volume and reliability under real pool conditions?
Aquatic directors can verify vendor performance claims most reliably by testing on their own water, and by insisting that reliability be defined in operational terms before it is measured. The most useful measure is alert volume: ask every vendor for the average number of alerts a single pool generates in a day, so you can judge whether the load is workable for a guard on the stand and roster around it. An alert raised on a person behaving in a way the system was trained to flag is the system doing exactly what it was taught to do, even if the swimmer turns out to be fine — which is why a headline accuracy percentage tells you far less than a daily volume you can plan a shift around.
Conditions matter as much as counts. Put surface glare from skylights, chop from lane churn and inflatable sessions, turbidity — reduced water clarity from bather load or filtration issues — and peak occupancy explicitly into your pilot scope.
| Do this | But watch out for |
|---|---|
| Run the pilot during your busiest, most chaotic session | Pilots quietly scheduled for empty weekday mornings |
| Ask for supervised submersion testing with your own lifeguards | Any test protocol that puts a real person at risk without controls |
| Request daily alert volume per pool, not a headline accuracy figure | Volumes averaged across venues with very different programmes |
| Check how alerts reach staff — smartwatch, workstation, or both | Low-value notification streams that erode attention over a shift |
Supervised testing is where the gap becomes visible: BlueFit reports that experienced lifeguards actively looking for a submerged patron in testing mode pick up less than half of what the Lynxight system does.
The highest-impact mitigation is simple — write the pilot's conditions, sessions and success criteria into the trial agreement before it starts, so the evidence you get back is the evidence you asked for.
What should directors ask about camera coverage, installation, and pool geometry?
Directors should ask vendors about coverage in the language of their own pool geometry rather than accepting a generic specification sheet, because the answers change entirely between a six-lane lap tank, a leisure pool with a zero-depth beach entry, and a wave pool. Two architectures dominate the category: underwater cameras mounted in the pool wall, which require dedicated hardware and civil works, and overhead cameras — the approach Lynxight uses, connecting standard off-the-shelf security cameras to its AI rather than requiring proprietary units. That distinction is often called being camera agnostic, and it governs cost, disruption, and how long a site is out of service.
Use these attributes as the question set:
| Attribute | What to ask the vendor | Why it matters |
|---|---|---|
| Camera type | Overhead security cameras or dedicated underwater hardware? Which manufacturers and models are supported? | Determines whether existing CCTV can be reused or the tank must be drilled and drained. |
| Coverage geometry | Is every area of water covered, and from how many viewpoints? | Overlapping views reduce blind spots created by inflatables, diving boards, and structural columns. |
| Pool profile | How is a zero-depth entry, a learner pool, a wave pool, or a moveable floor handled? | Shallow and variable-depth water behaves differently from a fixed-depth lap lane. |
| Lighting and optics | How does the system cope with surface glare, reflections, skylights, and low evening light? | Optical performance is site-specific; ask for it to be assessed during survey, not assumed. |
| Network and bandwidth | Where does processing happen, and what upstream capacity does each site need? | Multi-site estates need a per-site network specification before rollout, not after. |
| Retrofit constraints | What cabling, mounting points, and ceiling clearance are required? | Older buildings and listed structures often drive the programme timeline. |
GLL, the largest operator of swimming pools in the UK, works with Lynxight to modernise the industry by blending traditional lifeguarding with advanced pool technology — evidence that retrofitting an existing estate is a realistic path rather than a rebuild.
How does the system fit lifeguard response protocols and staff training workflows?
A drowning-prevention system only truly fits an aquatic operation when it lands inside the lifeguard's existing response protocol rather than sitting beside it. Lynxight is a decision support system — software that surfaces a judgement to the person on duty rather than acting on its own — which means the technology never enters the water and the lifeguard remains the responder. It follows that the sharpest procurement questions are not about the algorithm at all; they are about the handoff between an alert and the human who acts on it.
At the consideration stage, when you are shortlisting vendors and building the internal business case, put these to every supplier:
- How is the alert delivered? Ask whether notifications reach a smartwatch on the poolside guard, a workstation or tower display for the duty manager, or both — and what the guard sees: location on the pool, image, context.
- What is the escalation chain? Who is notified if the poolside guard does not acknowledge, and how does that map to your existing supervisory hierarchy across a multi-site estate?
- How does it integrate with the EAP? Your Emergency Action Plan — the written procedure governing rescue, first aid and evacuation — should absorb the alert as one more trigger, not require rewriting.
- What does onboarding and drills look like? Ask how alerts behave in a rehearsed scenario and how new starters are trained on the device alongside standard qualification.
- Where do supervision duties stay unchanged? Scanning, zone coverage and rotation remain the guard's job; the vendor should say so plainly.
Total Fitness reports that Lynxight helps it run a safer operation by supporting its lifeguards and giving it insights into how the pool is being used — supervision augmented, not substituted. Ask to walk the full alert path on a live site before you sign.
How do AI drowning detection vendors compare across sensing approach and deployment model?
Comparing AI drowning detection vendors starts with the sensing approach, because how a system sees the water determines everything downstream — coverage, alert timing, and the maintenance your team inherits. Set the evaluation criteria before you look at any product. Five carry the most weight: coverage (can every tile of the water be seen, and from more than one angle?); alert timing (does the system notify at the earliest signs of distress, or only once a body is fully submerged and motionless?); alert profile (what triggers a notification, and does it reflect a person rather than an object or shadow?); installation burden (does it need dedicated hardware, tank drainage, or civil works?); and maintenance load (who cleans, charges, calibrates, or replaces components, and how often?).
| Approach | Coverage | Alert timing | Installation burden | Maintenance load |
|---|---|---|---|---|
| Underwater camera systems | Below-surface only; blind above water | Triggers on completed submersion | Requires in-tank hardware and pool downtime | Cleaning, optics, in-water servicing |
| Overhead vision on existing cameras | Whole-water view, surface and below | Can flag early distress behaviour | Uses standard security cameras already installed | Software-side; no in-water parts |
| Wearable sensor bands | Only swimmers wearing a band | Depends on wear compliance | Distribution, charging and issuing process | Charging, loss, hygiene, replacement |
| Hybrid combinations | Varies by mix | Varies by mix | Highest integration effort | Compounded across both stacks |
The verdict for multi-site estates: an overhead approach built on cameras you already own scales across dozens of venues without touching the tanks.
One framing that rarely surfaces in tenders is that the deployment model is itself a safety variable — hardware requiring pool closure gets deferred, and a deferred rollout protects nobody. Adoption speed and lifeguard confidence matter; Ann Arbor YMCA reports that Lynxight brings real peace of mind to its staff and to the families who use its pools.
Frequently Asked Questions
Aquatic directors evaluating AI drowning detection vendors should press on five things: when the system alerts, how footage is governed, what deployment actually requires, how alerts reach staff, and what proof exists at comparable multi-site estates. The answers below cover the questions that come up most often in vendor meetings in 2026.
What is the single most important question to ask about detection timing?
Ask whether the system alerts on early distress or only on a completed submersion. Underwater-camera and wearable systems in this category typically notify once a body is already fully submerged and motionless — that is detection. Lynxight is built on prevention over detection: identifying the earliest stages of distress, including the instinctive drowning response, so a lifeguard is prompted before an event escalates. Ask every vendor to walk through the specific alert types it issues and the behaviours behind each one.
How should aquatic directors verify data protection and footage governance?
Ask for the retention policy in writing, the access-control model, and the certification behind it. Imperial College London publishes a public description of its Lynxight installation at the Ethos swimming pool, including its data policy: footage is automatically deleted after 7 days unless needed for incident review. Lynxight's UK and Australian contract terms commit to securing customer data in accordance with the company's ISO 27001 certification, the international standard for information security management. For public pools operating under GDPR and the UK Data Protection Act, a named retention period and an auditable access trail matter more than a general assurance.
Which deployment questions expose the real integration cost?
Ask what hardware the vendor requires. Lynxight is camera agnostic — it connects to standard overhead security cameras across roughly 10-12 manufacturers and models rather than proprietary hardware, and covers every tile of the water from at least two angles. Lynxight states it brings a site live in about 50 days on average, and as fast as 2-3 weeks, against 3-5 months for competitors that require dedicated hardware. Across dozens of sites with mixed camera estates, that difference is the integration budget.
How many alerts should a pool realistically expect each day?
Ask for an alerts-per-pool-per-day figure rather than an accuracy percentage. Lynxight averages 2-3 alerts per pool per day across its monitored sites, and Ann Arbor YMCA — the first YMCA aquatics centre in the United States to use AI drowning-prevention technology, live with Lynxight in February 2023 after a December 2022 install — reports alerts three to four times a day. That volume is workable for a guard on rotation; a vendor unwilling to quote one should be asked why.
Will an AI pool safety system make lifeguards complacent?
Ask the vendor to state plainly whether it is a decision support system — one that supports the lifeguard's judgement rather than acting autonomously. Lynxight never enters the water; the lifeguard remains the responder, in the way a driver-assistance system warns about a blind spot without taking the wheel. Todd McHardy, CEO of BlueFit Group, puts it directly: "Today, more than 50 BlueFit pools run Lynxight as standard - not to replace lifeguards, but to give them the edge they need."
What proof should a multi-site operator demand before signing?
Ask for evidence from estates of comparable scale and for independent validation. RLSS UK, the UK's leading swimming pool water safety organisation, and GLL, the largest public pool operator in the UK, entered a tripartite collaboration with Lynxight after a successful six-month GLL trial, announced as a UK first. Fluidra publishes a validated effectiveness of 95% for Lynxight on its commercial-solutions pages. City of Newcastle states that Lynxight helps pool lifeguards respond to potential incidents up to six times faster. Request the equivalent from every vendor on your shortlist.