The short answer: the $899 Litter-Robot 5 Pro scoops waste well and its AI does not work, and that is the finding of two professional tests run six months apart on the same machine. The Verge spent six months with the 5 Pro and with the cheaper $599 Evo, and its reviewer wrote that she “wouldn’t recommend paying the premium for the 5 Pro’s AI features” because they “just don’t work”. BGR, who had run an LR4 for years before testing the Pro, came to the same place from the other direction and called the newer machine “a bit of a step backward when it comes to reliability”.

Cats.com ran a third test, three weeks in April with two cats, and scored the box 4.6 out of 5 overall with price at 3 out of 5. Its reviewer could not test the two headline features at all. WasteID had not shipped, and the AI facial recognition that assigns visits to a cat needed a membership the review did not have. That is the shape of this product: excellent at the job a litter box has always done, unfinished at the job that costs $200 more than the model below it.

petiqlab read all three tests, Whisker’s product page for the 5 Pro and the brand’s Trustpilot record on October 4, 2026, and has not tested this machine. What follows is what the reviewers with months of hands-on time published, what the subscription actually covers and when each piece arrived, and the checks worth running before AI is the reason you pay $899 instead of $699.

Three testers, one machine, the same complaints

Tester	Time with the 5 Pro	What they reported
The Verge, published August 17, 2026	6 months, alongside the $599 Evo	AI features “just don’t work”, the $8-a-month plan gave no actionable insight, Cat ID merged two cats, WasteID missed repeatedly, only the front camera records
BGR, published February 2026	Months, after years on an LR4	Weekly false “extended scale activity” alerts, regular motor issues, app reporting the wrong state, a setup process that “crashed multiple times”
Cats.com, published April 4, 2026	3 weeks with two cats	4.6 out of 5 overall with price at 3 out of 5, WasteID not released, facial recognition untestable without a membership

The order of those dates is the part worth sitting with. BGR was writing about reliability problems in February. Cats.com found the flagship features unshipped in April. The Verge still found them broken in August. Nothing in the three reports suggests a launch-week bug that a patch fixed, and the product page’s own FAQ explains why the wait is baked in. Whisker writes that “features become more accurate over time as they learn from your household, typically optimizing after 30 days of regular use”. That is a fair sentence for a machine that learns. It is a poor apology for one that shipped.

Cat ID merged two cats into one

The identification feature is the hinge on a multi-cat box, because it decides which cat owns each recorded visit, weight and trend. The Verge gave it six months and two cats that look nothing alike, one large gray and white tabby and one small tuxedo. The cameras “couldn’t tell” them apart, “getting confused between them and eventually merging them into two identical cats”. The visible symptom was worse than a label swap. Her 13.4-pound cat lost 2 pounds overnight on paper, with the weight recorded under his brother’s profile, and a week later both cats still registered the same weight.

Whisker’s explanation, as quoted in that review, is that poor lighting and bad enrollment images throw the camera off, and that when the system is unsure it uses other signals such as the onboard scale to guess. The company told the reviewer that the unit’s feces-sensing gas sensor was off, then that the scale was off, which “would require a whole new machine to fix”. It also said an update made enrollment “more selective about the images it uses for facial recognition”, and that the system is built to learn.

Two details from the same test are more useful than the technical explanation. Whisker sells a $29 RFID PetTag as the answer for cats that look similar, and the reviewer’s verdict on that was blunt: “A $900 machine advertising facial recognition shouldn’t need a wearable to tell two different-looking cats apart.” The other detail is that the box also reads weight low in general. The Verge’s line about it is the one to remember before trusting an app with a health claim, since it noted that the robot still thinks the cat is lighter than he is.

BGR’s report describes the same sensor layer failing in a quieter way. Motor issues “popped up regularly even though the unit wasn’t overfilled”, the sensors “kept detecting ‘extended scale activity’ when no cat was anywhere near the robot”, and the false alerts arrived “at least once a week during testing”. The app then reported the wrong thing often enough that the tester stopped trusting it as a monitor, including “low litter warnings when there was plenty and offline notifications when the unit was connected”. An automatic box that needs a weekly manual reset is a different product from the one in the advert.

What $7.99 a month buys, and when each piece arrived

The subscription is where the AI story gets priced. Cats.com read the tiers for this review and reported that Whisker+ runs at $7.99 per month or $79.99 per year. Without it you keep the past two days of video recordings and up to 5 minutes of live streaming per day. A membership adds unlimited live streaming, 30 days of recordings, long-term trends of up to 2 years, anomaly detection and PetTag compatibility. The Verge rounds the same plan to $8 a month, and the arithmetic that matters is simple. Six months of testing is roughly $48 of subscription on a machine that, in that reviewer’s words, “didn’t deliver one actionable or useful insight”.

The feature timeline is the part a buyer should price in:

Feature	State when reviewers tested it
WasteID, the sensor that separates urine from stool	Not released in April 2026 per cats.com, which could not test it; still “repeatedly failed to accurately identify waste” in August per The Verge
Cat ID facial recognition	Requires Whisker+; merged two visually different cats after six months; up to 5 cats advertised
Interior camera recording	Live view only in August; Whisker told The Verge the interior recording is “coming soon”
Patterns, the daily summary	Arrived several months after launch, so owners “have been paying $8 a month for features that didn’t work”
Long-term trends, anomaly detection	Membership only, up to 2 years of data

Whisker’s own product page describes the hardware as a “dual AI-powered camera system” that allows “real-time streaming, recording, and cat identification using AI-powered facial recognition, designed for homes with up to 5 cats”, and it answers more questions than the advertising does. The same page carries the 30-day learning note, the parenthetical “(Some features require Whisker+.)”, a 90-day in-home trial and free shipping, and it lists the specs. The box stands 29.9 inches tall with a 22 by 27 inch footprint, a 15.75 inch entryway and 17.5 inches of interior height. It weighs 29.6 pounds empty and is rated for cats of 3 to 30 pounds. Those numbers are why this box works for large cats in a way a $200 machine does not, and the size question is covered in more detail in the guide to automatic boxes for large cats.

The trial length is the most useful number on that page. Whisker’s own FAQ says the learning features optimize “after 30 days of regular use”, and the return window is 90 days. That leaves two months to test the AI for yourself before the money becomes final, which is a longer runway than most pet tech gives you.

The $200 question

Four Litter-Robots were on sale on October 4, 2026, and the prices sit close enough together that the choice is about trade-offs rather than budget. These are the manufacturer’s own listings, with the review counts shown on each page:

Model	List price	Site rating	What the extra money buys, and what testers said
Litter-Robot 5 Pro	$899	4.4 from 302 reviews	Two cameras, Cat ID, WasteID. The AI was the least finished part of the machine in every long test
Litter-Robot 5	$799	4.6 from 241 reviews	No camera, keeps WasteID. The Verge’s recommendation: “save yourself $100”
Litter-Robot 4	$699	4.5 from 14,096 reviews	No AI at all. The Verge called it excellent without it, and BGR found it more reliable than the Pro
Litter-Robot EVO	$599	4.7 from 35 reviews	Smaller bin, no litter-level monitoring, no night light, no hopper compatibility. The Verge: not worth the $100 saving over the LR4

Two lines in that table are worth reading twice. The LR4 has 14,096 reviews on Whisker’s own page against 35 for the Evo, which tells you where the installed base actually is. The highest score in the table also belongs to the machine with the fewest features the company is trying to sell you. That fits the pattern in every long test. The less the software has to do, the better the machine behaves.

petiqlab’s position on these four is on record. The 5 Pro is rated a disappointment in this site’s review of it, the LR4 stays neutral rather than recommended, and the Evo is treated as a first-year model to avoid in the Evo write-up. Nothing in the professional tests above changes that, and the LR4’s own long-term costs are documented in the two-year ownership breakdown.

Where the camera ends up pointing

A camera on a litter box is also a camera in your home, and the most specific warning in the Verge test took a week to notice. The front-facing camera catches cats on the way in, and in that reviewer’s placement it also recorded anyone using the powder room across the hall. Turning on the setting that records only when a cat is detected does not fix that. As the reviewer noted, cats like to follow their owners into the bathroom, which can put both of them in frame.

The privacy controls that do exist are worth knowing before the box arrives. Without a membership, recordings last two days, which limits what a miscalculated placement leaves behind. With one, they last 30 days. On placement, the practical rule from that test is to check what is behind the box as carefully as what is in front of it, since the camera looks out past the globe rather than down into it. Behaviour of the indicator light on this brand’s cameras is its own topic, and the write-up on the Petlibro camera light covers why a light that stays on is not always proof that the camera is off either.

Five checks before you pay for the AI tier

1. Ask what works on the day you buy. In April, WasteID had not shipped. In August, the interior camera still only streamed. A feature list with “coming soon” in it is a promise, and the Verge’s general rule after six months is worth repeating: “never buy a product based on the promised future updates”.
2. Run the identification test inside the trial, not after it. Whisker’s own FAQ gives the learning features 30 days to optimize, and the trial runs 90. Put both cats through, then open the app’s history and check whether each visit is attached to the right animal. If two cats have merged, that is the answer, and it arrived inside the window where the money can still come back.
3. Price the plan into year one before comparing models. The gap between the $699 LR4 and the $899 5 Pro is $200, and the plan that unlocks the Pro’s health insights is another $79.99 a year if paid annually. Without the plan, the recordings fall to two days and the trends stop.
4. Count the manual resets the automation adds. BGR’s weekly false scale alerts and regular motor faults turned a hands-off machine into one that needed attention roughly every week. A $899 box that quietly asks for ten minutes a week is still a better deal than scooping twice a day, as long as you priced the ten minutes in.
5. Check whether software is the right tool for the thing you want. If per-cat data is the goal, the category already solves identification with hardware. A microchip reader in a feeder cannot confuse two cats and needs no account, which is the argument in the microchip feeder guide, and the counting logic for multi-cat homes is in the litter box number guide.

What to buy instead

No brand in this category has published a per-cat identification accuracy figure that this site can verify, and the 5 Pro is not an outlier so much as the most expensive example of the problem. The Verge and BGR both landed on the same practical answer, which is to buy the generation without the AI premium, or buy the tier that does the mechanical job and skip the insights.

Whisker’s own posture is worth putting beside the reviews. Trustpilot showed litter-robot.com at 3.3 stars from 606 reviews on October 4, 2026, with the profile reporting that the company replied to 96% of negative reviews, typically within 1 week. That is a strong reply rate, and it is not the same as a fix, a distinction this site unpacked in its comparison of complaint platforms. What the reply rate cannot tell you is how many of those replies ended with a working machine.

For a household that wants automation without the AI tier, petiqlab’s standing pick in this category is the PETKIT Purobot Max Pro 2 (B0DFYLFFG1 on Amazon), listed at $499.99 on PETKIT’s own store, with a two-year warranty and no subscription attached to the machine. Three disclosures belong with that recommendation. The camera version’s clip recording runs for a trial period and then drops to live view only, so buy it for the drum and not the camera. One owner has reported a cat urinating onto the camera opening on this model. And its own identification feature is labelled beta, with owners reporting it misses visits often enough that it should not be trusted as a health record, which is covered in the multi-cat guide. If the identification failure described above is why you are reading this, the honest sequence is a microchip feeder for the per-cat data and a plain high-sided box, or a box whose job is scooping, before it is an AI system.

For watching the cat rather than the box, the cheaper route is a camera that records locally. The TP-Link Tapo write-up covers a $35 to $40 camera that stores video on a microSD card, and the arithmetic of choosing local storage over a monthly plan is in the subscription cost comparison. One camera at $40 with no plan costs less than five months of Whisker+, and it will not mislabel which cat walked past it.

FAQ

Does the Litter-Robot 5 Pro need a subscription? For the features most buyers want the Pro for, yes. Without Whisker+ you keep the past two days of recordings and up to 5 minutes of live streaming a day. Whisker+ is listed at $7.99 per month or $79.99 per year and adds unlimited live streaming, 30 days of recordings, trends up to 2 years, anomaly detection and PetTag compatibility. The product page also states that some features require Whisker+.

Is the AI still broken now? The most recent long test available is The Verge’s from August 17, 2026, after six months of use, and it found Cat ID merging two visually different cats, WasteID failing to identify waste, and interior recording still unavailable. Whisker says the system learns and that enrollment was made more selective. petiqlab has not tested the machine and cannot confirm whether a later update changed any of this.

How much does the whole thing cost in the first year? The 5 Pro listed at $899 on October 4, 2026 with a 90-day in-home trial and free shipping, and Whisker+ adds $79.99 if paid annually. That is $978.99 for the first year, before litter, liners, the $29 PetTag or the $12 carbon filters, and $899 plus monthly fees in any year after if the plan continues.

Do both cameras record video? No. The front-facing camera records, and the camera that looks into the globe is live view only, according to The Verge, which also reported that Whisker described the interior recording as coming soon. Cats.com reported the same limitation in April. The advertising for the machine implies recordings from both.

Which model should I buy? The reviewers’ advice and this site’s position point the same way on the Pro, this one is not worth the AI premium. Between the others, The Verge suggested the $799 Litter-Robot 5 if you want WasteID without the camera, and called the $699 LR4 excellent without any AI, while BGR said the LR4 might be the better pick on reliability. The $599 Evo was judged not worth its saving by The Verge, and this site treats it as a model to avoid.

If identification is the point, what actually works? Hardware, as far as this site can verify. A microchip feeder reads the chip implanted in each cat, runs on no account and no firmware, and cannot merge two animals. It solves feeding rather than litter, and for litter the choice is between a box without AI and an honest schedule.

Does a wrong identification matter that much? It changes whose health record you are reading. In the Verge test a 13.4-pound cat briefly weighed 2 pounds less and had that weight filed under his brother for a week, and both cats read the same weight. Trends and anomaly alerts are only as good as the assignment behind them, which is the finding in this site’s audit of AI health alerts and the reason health data behind paywalls deserves a separate look before you rely on it.