Skip to main content

How works

About What AI Would Buy

Everyone does the same homework before they buy. Almost no one finishes it.

What AI Would Buy is an independent weekly chart of the products and brands recommended by leading AI assistants. Every leader is cross-checked against expert tests, owner reviews, professional press, marketing claims and safety data — so buyers can distinguish what is visible from what is actually worth buying.

Ask an AI, read the reviews, watch the videos, then decide. It’s the right way to choose — and it takes hours. So most people quit after the first step, trust one opinion, and hope. We run the whole thing for you, properly, and show our work.

01

The long way — do it yourself

four steps · one opinion · no audit trail
STEP 01~5 min

Ask an AI

“What’s the best robot vacuum under $600?” One answer, confidently written.

One model, one take. You can’t tell if it’s real or invented.
STEP 02~45 min

Read the reviews

Pro reviews, retailer blurbs, a dozen Reddit threads — half of them sponsored.

They contradict each other. Now you’re refereeing strangers.
STEP 03~90 min

Watch the videos

Four or five YouTube teardowns at 1.5×, skipping the sponsor reads.

This is where people quit. The patience runs out before the answer.
STEP 04~20 min

Decide

Reconcile everything in your head and finally commit to one.

Still second-guessing the moment you add to cart.
~2.5–3 hrs

That’s the cost of doing it right — once, for one product. Almost nobody pays it. They stop at step one, take a single AI’s word, and never find out the models disagreed.

We do all four jobs — for every product — before you arrive.

02

The short way — already done

three juries · cross-checked · every source shown
The same four jobs
Asked all six AIsChatGPT · Claude · Gemini & others — not one, all of them
Read the criticsPro reviews scored and summarised
Watched the verdictVideo reviews distilled to a score spread
Weighed the buyersThousands of real owner reviews
~2 minto read what took us the week to assemble
Example verdict
RRoborock S8 ProROBOT VACUUMS · UNDER $600Trust
AI consensus6 of 6 agree#1
Critics14 pro reviews4.3
Buyers2,140 owners4.5

All three juries point the same way — that’s why it’s a Trust. When they split, you’d see a Check instead, and exactly who disagreed.

03

The difference, to scale

same research · same decision · real proportions
Youdo it yourself
Ask5m
Read reviews45m
Watch videos90m
Decide20m
~3 hrs
Uswhat ai would buy
everything above — already assembled, cross-checked, on one screen
~2 min
≈3 hrshanded back every time you’d otherwise run the search yourself — and you skip the part where you quietly settle for one opinion.
The real point

Speed is just the by-product. What you actually get is the rigorous version of the search — the one you’d never sit through yourself.

01

Not faster — finished.

The hours are real, but the bigger win is the work you’d never actually do. We always complete all four steps. You almost never would.

02

Six AIs, not one guess.

A single model can hallucinate a confident answer. We ask all six and only call it settled when they agree — the one thing you can’t replicate in a single chat.

03

You can check our work.

Every verdict shows which AI said what, what the critics scored, and what owners reported. No black box — disagreement is on the page, not hidden.

04

Who reads this

same board · four different questions
Shoppers

Pick the right one without becoming an expert first.

Browse the chart
Brand teams

See whether the assistants name you — and where they name someone else.

Check a brand
Analysts

Watch a category move week over week without running the queries yourself.

This week’s moves
Investors

Read mindshare as a signal: who the models started recommending, and when.

The brand chart

Do the three hours yourself — or read the answer and spend them on something else.

Find your product
The fine print
  • Nobody pays to appear here. No sponsored placements, no paid inclusion, no arrangement that moves a rank. The board is whatever the assistants said this week.
  • A high rank is visibility, not a verdict. It means the assistants are likely to name that product — which is not the same as it being the best one. How the rank is computed.
  • The panel is 6 assistants this week, and the lineup changes as the field does. Who is on it right now.
How we actually verify it
See the method

Weekly email

The week's AI picks — in your inbox.

  • Each model's #1 — and where they disagree
  • The biggest movers & newcomers
  • Reviewer & buyer reality checks

One email a week · first issue coming soon · unsubscribe any time.