How to use AI to compare products before you buy

Write down what you actually need

Most bad purchases are a needs problem, not a product problem. Let the AI interview you before anyone mentions a brand.

I'm shopping for a [product type, e.g. robot vacuum]. Before suggesting any products, ask me up to 6 questions, one at a time, about how I'll use it, my space, my budget, what I disliked about past ones and any deal-breakers.
Then summarize my needs as: Must-have | Nice-to-have | Don't care | Budget ceiling.

Learn which specs matter

Every category has a few specs that matter and many that are marketing. Ask the AI to explain them for your use, not to rank brands.

For my needs above, explain the 5-7 specs or features that matter most when comparing [product type], in plain English.
For each: what it means, what level is enough for me, and what's mostly marketing.
Also list common trade-offs (e.g. battery life vs weight) and typical hidden costs like refills, subscriptions or accessories.
Don't recommend specific models yet.

Build a comparison table from real sources

Pick 2 to 4 candidates from wherever you trust: a store shelf, a review site, friends. Copy each one's specs from the maker's official product page or the retailer listing and paste them in. Keep the exact model number; many brands sell near-identical names with different specs.

Here are the official specs for [N] products, copied from their product pages:
--- Product A: [exact model number] ---
[paste specs]
--- Product B: [exact model number] ---
[paste specs]
Build a comparison table using ONLY the text above. Columns: each product. Rows: the specs that matter for my needs, plus price and warranty.
If a spec is missing for a product, write "not listed". Do not fill gaps from memory.
Below the table, say which product fits my Must-haves best and why, in 3 bullets.

The "not listed" rule is the important part. A gap in the table is a question to check, not a number for the AI to guess.

Find patterns in real reviews

Copy 15 to 30 reviews from more than one place, and include the 2- and 3-star ones, which tend to be the most specific. Then ask for patterns, not a verdict.

Here are customer reviews of [model], copied from [sources]:
[paste reviews]
Group the complaints and the praise into themes. For each theme: how many reviews mention it, and one short quote.
Flag problems that show up after weeks or months of use (failures, battery fade, customer service).
Point out reviews that look unhelpful or suspicious, e.g. vague praise with no details or many near-identical wording.
Don't give an overall star rating.

Review counts from a sample you pasted aren't statistics, but repeated, specific complaints ("the hinge cracked after 3 months") are worth taking seriously.

Compare the real cost, not the sticker price

Estimate the 3-year cost of ownership for each product using the prices I give you:
Product A: price $[ ], [refills/filters/ink] $[ ] every [period], subscription $[ ]/month (if any)
Product B: ...
Show the math as a calculation for each line, then a total. Mark anything you assumed.

Re-check the totals in a calculator or spreadsheet; chatbots can slip on arithmetic. Our monthly budget guide shows how to fit a big purchase into your plan.

Argue against your favorite, then verify

Once you're leaning one way, ask the AI to make the case against it. It is good at surfacing what you were glossing over.

I'm leaning toward [product] because [reasons]. Play devil's advocate: give me the 3 strongest reasons this could be the wrong choice for me, and what I'd regret in 6 months.
Then give me a short checklist to confirm before I buy: exact model number, current price at 2-3 retailers, return window and who pays return shipping, warranty length and how claims work.

Do that checklist on the retailers' own pages, not in the chat. If your AI tool shows shopping results or product cards, treat them like any other listing: check the price and model number on the store page and whether anything is labeled sponsored. For research that leans on articles and claims, our AI research and fact-checking guide covers how to trace claims to sources.