Which tool to choose
They're not equivalent. Choose based on what you need to compare.
For product reviews, hotels, software, purchase decisions: Perplexity. It's built for this. The difference with ChatGPT is sharp: Perplexity was born for research with cited sources, ChatGPT is at its best for writing, long drafts and code. The concrete lever is Focus mode, the mode that narrows the search to one type of source: you can limit it to Academic, YouTube or Reddit. Reddit is gold for the opinions of people who actually use a product.
For a written report you put on record (e.g. a comparison of vendors for work): Gemini with Deep Research, even for free. That's the concrete advantage over ChatGPT, where Deep Research requires a subscription. The limit to know before starting: the free plan has a cap of 5 Deep Research reports a month. Save them for the comparisons that matter.
If you already use ChatGPT and don't want to switch apps: it's fine anyway. Web browsing also works on the free plan. For the occasional review comparison it's enough.
How to do it
The method is identical on all three tools, and doesn't change between browser and app: what counts is how you phrase the request. A generic question gives you a generic summary. You have to ask for the structured comparison.
- Open the tool you chose.
- On Perplexity, turn on Focus mode before writing: the mode icon is near the bar where you type; select Reddit for real opinions or Web for general reviews. On ChatGPT, enable web search from the tools menu next to the input bar.
- Paste the structured request below, changing only the subject of the comparison.
- Read the answer and click the notes: for important decisions, always verify what the cited sources actually contain.
- Do a follow-up on the weak points that emerged (e.g. "dig deeper into the complaints about customer support").
The operational syntax for the comparison:
Compare the online opinions on [PRODUCT/SERVICE A] and [PRODUCT/SERVICE B].
Organize the answer into three blocks:
1. Recurring pros of each, with how many different sources cite them
2. Recurring cons and complaints of each
3. The points on which the reviews contradict each other
Use only real reviews and discussions from the last 12 months.
For each statement indicate the source. Don't invent numbers or quotes:
if a piece of data isn't verifiable, say so.
Feedback: a good answer gives you at least 4-5 different sources and flags where opinions diverge. If it gives you back three vague sentences with no links, web search wasn't on. Go back to step 2.
Concrete example
You have to choose between two robot vacuums of about 300 euros and the reviews on the e-commerce sites all seem five-star, suspiciously so.
You open Perplexity, set the Focus on Reddit and paste the prompt above with the two exact models. In a few seconds the answer arrives: model A gathers praise for suction power, but three different threads report that the app disconnects from WiFi after a few weeks; model B has more lukewarm reviews on suction, no recurring complaints about the software, and one user recounts that support replaced the brush without arguing. You click the notes: two lead to real Reddit discussions, one to an industry forum. You ask as a follow-up "how many users cite model A's WiFi problem?" and get the detail. You choose B, not because it suctions better, but because A's flaw is the one that would have driven you crazy. Decision made in three minutes, with the evidence in hand.
When it does NOT work (and how to fix it)
If the AI invents reviews or numbers
It happens especially when web search isn't on and the model answers "from memory". The way out: demand the source for every statement (it's already in the prompt above) and click the links. If a link doesn't exist or doesn't contain what the AI claims, that line is made up: discard it. On Perplexity the risk is lower, because every sentence is anchored to a citation you can open.
If you've run out of the day's free comparisons
On Perplexity's free plan the richer comparisons (the Pro searches) run out fast, around 5 a day. When you finish them, two routes: switch to free ChatGPT, where basic web search stays available, or use Gemini bearing in mind the cap of 5 monthly Deep Research reports. By spreading the comparisons across the three tools you never run dry in a single day.
If the reviews it cites are old or for a different product
The AI sometimes mixes up models with similar names or pulls reviews from three years ago. Fix: write in the prompt the full name with year and model code, and repeat "last 12 months". If it gets it wrong again, paste the links of the specific pages yourself and ask it to analyze only those.
If you need the decision put down in writing for others
The quick chat isn't enough when you have to justify the choice to a boss or a client. Here you move to a real report: Gemini Deep Research browses the web autonomously, synthesizes multiple sources and produces a structured, cited document (free, within the 5 monthly). It's what you attach to an email without blushing.
A tip from someone who actually uses it
Never ask "which is the best". It'll give you a diplomatic, useless answer. Instead ask "what are the recurring complaints" and "what do the reviews contradict each other on". Flaws repeated by different sources are the most reliable signal: one isolated star can be an angry user, but the same problem in five different threads is a fact. For big decisions, cross-reference at least two tools: what one leaves out, the other often catches.
Frequently asked questions
Can I trust the reviews the AI summarizes for me?
Trust the method, not the words. The AI is excellent at gathering and grouping scattered opinions, but the verification stays yours: the clickable links are there for this. Treat the summary as a starting point, not as the final verdict.
What's the fastest way to get real opinions and not marketing?
Perplexity with the Focus on Reddit. You cut out the affiliate sites that exist only to push you to buy and go straight to people who own the product. A specific question like "best CRM for a 5-person sales team under 50 euros a month" beats a generic "best CRM".
Do I have to pay to do good review comparisons?
No, for normal use the free plans hold up. Perplexity's is among the most generous. Paying makes sense only if you do comparisons every day for work: in that case you'll exhaust the free limits constantly.
Doesn't the AI just read the same reviews I'd find myself on Google?
Partly yes, but the value isn't finding the sources: it's cross-referencing them. By hand you'd have to open twenty tabs, read hundreds of comments and keep the patterns in mind. The AI reads those pages for you and puts the contradictions in front of you. And with the Focus on forums and communities it pulls out discussions you'd never see on the first page of Google, dominated by SEO and affiliations. The advantage isn't access, it's the time and the synthesis.