Most waymo reviews read like a passenger diary, full of adjectives but short on control. I wanted reproducible data. So I took my standardized AV test suite to Phoenix and ran it against Waymo's ride-hail service — nine rides, 23 scenarios, every disengagement logged. Here's what the robot actually did, what it should have done, and where the gap sits.
How I Set Up the Test
My rig is the same one I use for every production system I evaluate. No smartphone on the dash, no GoPro angles. I record the vehicle's reported state through the Waymo app telemetry, a dual-band GPS logger, and a LiDAR-equipped backpack that captures scene geometry. Each ride gets a severity rating from 1 to 5, where 5 is "a human driver would have honked." I also log every intervention: hands-on-wheel, brake override, or the moment I asked the remote operator to take over.
I ran the full suite in Phoenix in late March, covering the Sky Train stretch, the Scottsdale loop, and the downtown grid. Rides were random, not cherry-picked. I didn't tell the company I was testing. That's the only way to get unvarnished data.
Those are the facts behind the waymo reviews I publish — and the ones most YouTube reviewers skip.
Where the Robot Is Genuinely Good
Let me give credit where the data supports it. Waymo's lane-keeping and lateral control are excellent. In 14 of my 19 highway miles, the car stayed centered within six inches — better than most human drivers I share the road with in Michigan. Merges were decisive; the ride was stable enough that I almost forgot to watch the road (I didn't).
Left turns across traffic — historically the hardest maneuver for any autonomous system — succeeded 11 out of 12 times. That's a real accomplishment. Four years ago the same platform struggled with unprotected lefts in dense traffic. The improvement line is visible in the data, not just in press releases.
The Hallucination Log
Here's where my notes turn into a bug tracker. Severity 3: on a wide, empty boulevard at 10 PM, the Waymo slammed on the brakes for a shadow cast by a palm tree. The perception stack tagged it as a pedestrian for 1.2 seconds before reclassifying. That's a classic false positive — safer than a false negative, but it tells me the model still has a blind spot for high-contrast shadows.
Severity 4: approaching a construction zone on 7th Street, the car attempted a lane change into an occupied lane without a blinker. A human driver next to us laid on the horn. The safety driver intervened. That's exactly the kind of edge case that shows up in my standardized suite but rarely in a five-minute ride-along.
These are the incidents that never make it into the official waymo reviews you see on business sites.
Where a Human Driver Would Have Done Better
Let's talk about the moments that don't make the marketing videos. At a four-way stop in Scottsdale, the Waymo waited 14 seconds before creeping forward, even though there was no cross traffic. A human driver would have rolled through in three. That's not a safety violation, but it's the kind of overly cautious behavior that makes riders late and passengers queasy.
On the freeway, the car consistently brake-checked when a semi truck drifted toward its lane line, even though the truck had already begun moving back. The system treats every lane violation as an emergency. That's understandable, but it makes the ride less comfortable than a defensive human driver.
These are the judgment calls that separate "safe" from "smooth." My data puts Waymo at an 8 for safety but only a 6 for "human-like" decision-making.
The Scorecard
I scored Waymo against my baseline: a cautious but competent human driver. The aggregate results: 8.2 out of 10 for safety-adjacent behavior, 7.5 for comfort (the brakes are still abrupt at low speeds), and 6.0 for "does it make sense" — my catch-all for decisions that leave a passenger muttering.
Disengagement rate: two of nine rides required a human takeover. That's 22 percent, which sounds high, but every one was from the rider seat or a remote operator, not a safety driver grabbing the wheel. Compare that to early Cruise logs I've pulled, and Waymo is clearly ahead. But "clearly ahead" is not "almost solved."
Should You Trust the Hype?
The marketing says Waymo is the safest driver on the road. My waymo reviews add an asterisk: safest in the narrow scenarios the system has mastered, and still untested in the chaotic edge-case world that makes up ten percent of a normal commute. For an EV owner in Phoenix or San Francisco, the service is worth a ride. For an engineer, the telemetry is worth studying. For a passenger who wants a boring trip, it's close — but not there yet.
The Bottom Line
Waymo has turned a science project into a usable product. That's not nothing. But the gap between usable and unremarkable shows up in every brake-for-nothing event, every hesitant merge, every shadow mistaken for a person. My recommendation: take a ride, keep your hand near the imaginary brake, and read the raw logs before you call it a miracle.
I'll keep updating these waymo reviews as new OTA builds land. The data lives in a public Airtable, and I re-run the full suite after every update. If the robot improves, I'll show you the numbers. If it doesn't, I'll show you those too.