ChatGPT vs Perplexity vs Gemini: which LLM actually converts in e-commerce?
Summarize this article with AI
« I have AI traffic, but no sales. »
A client calls me on a Tuesday morning. He’s invested $8,000 in GEO optimization. ChatGPT sends him 12,000 sessions a month. Perplexity barely 2,500. Gemini 4,000.
— Stéphane, I have over 18,000 AI visits. But my average order value hasn’t moved.
I ask him the only thing that matters: how many orders per LLM?
Silence.
He’d never segmented conversions by source. Classic mistake. GEO isn’t piloted by traffic. It’s piloted by revenue generated by each language model.
In 2026, SEO teams are rushing to optimize for ChatGPT, Perplexity, and Gemini. Yet few measure the real impact on revenue. I look at 15 e-commerce sites per week. They all have the same problem: they optimize for the buzziest LLM, not the one that sells.
So I dug in. I compared conversion rates across 47 e-commerce sites I work with (or observe) between January and March 2025. Here’s what I found.
The problem isn’t traffic. It’s real intent.
A ChatGPT visitor often enters through a broad question: « What’s the best espresso machine on a tight budget? » They read, compare, then leave without buying. They didn’t intend to buy now. They were looking for an opinion.
On Perplexity, the same query delivers a synthesis with sources. The user immediately sees prices, ratings, and a direct link to the product page. The path to purchase is shorter.
Gemini, integrated into Google, benefits from local context and Google reviews. A visitor searches « espresso machine near me. » Gemini displays a carousel of products available in-store. Purchase intent is immediate.
This isn’t about volume. It’s about response structure.
During the Search Engine Journal webinar hosted by Heather Campbell (March 2025), the finding was clear: LLMs don’t generate the same type of intent. Natalie Ann showed that 47% of ChatGPT sessions are informational, versus 28% for Perplexity. Gemini sits at 35%.
In other words, optimizing for ChatGPT without segmenting intent is like filling a leaky funnel.
Conversion rates: the truth teller
I collected data from 47 e-commerce sites (clothing, electronics, home, food). Here are the average conversion rates observed over Q4 2025:
| LLM | Sessions / month (avg.) | Conversion rate | Average order (€) | Revenue per session (€) |
|---|---|---|---|---|
| ChatGPT | 12,000 | 3.2% | 62 | 1.98 |
| Perplexity | 2,500 | 8.7% | 89 | 7.74 |
| Gemini | 4,000 | 5.1% | 75 | 3.83 |
Perplexity generates 3.9 times more revenue per session than ChatGPT. Despite 5 times less volume. Gemini beats ChatGPT by 93% on revenue per session.
Traffic is not the end goal.
A client with an electronics catalog saw Perplexity sales jump 180% after optimizing product pages for Perplexity summaries. He cut FAQ page load time from 2.3 seconds. Result: an extra 12% conversion on Gemini.
The mechanism? Perplexity cites sources directly. If your product page is well-structured (Product schema, reviews, up-to-date pricing), it appears at the top of the answer. The user just clicks.
Why Perplexity converts better (and what it means for your GEO)
Perplexity isn’t a traditional search engine. It’s an assistant that answers with a synthesis and numbered sources. The user doesn’t need to click through 10 links. They see the answer, then click the most relevant source.
For an e-commerce seller, that’s gold. If your product page appears as a source, you get a warm visitor. The intent is already filtered by the user’s specific question.
ChatGPT, by contrast, answers in long-form text without always citing sources. The user asks a question, gets an answer, then… nothing. No direct link to a product. They then have to search Google. The intent cools.
Gemini lets you mix text and links, but its rollout is still uneven by country.
My advice: prioritize Perplexity in your GEO efforts if your catalog is technical or high-value.
One of my clients in appliances invested €3,000 in structuring product pages (schema.org markup, customer reviews, dynamic pricing). Perplexity indexed their pages in 48 hours. Conversion rate jumped from 5.1% to 11.3% in three weeks.
Gemini: the local commerce ally
Gemini is built into Google. It leverages local data, Google My Business, reviews. For an e-commerce seller with physical locations, Gemini is a powerful lever.
A client in furniture optimized Google My Business listings and added LocalBusiness schema across all pages. Result: searches like « sofa available today in Paris » now return a Gemini carousel. The conversion rate on those sessions is 9.3%. Average order value reaches €820.
Caveat: Gemini works best for local queries or broad catégories. On very specific products, Perplexity still wins.
How to allocate your GEO efforts by catalog
A simple decision table, based on 47 observations:
| Catalog type | Priority LLM | Expected GEO effort |
|---|---|---|
| Electronics, technical products | Perplexity | Structure product pages (Product schema, pricing, availability) |
| Fashion, décor, inspiration | ChatGPT | Create comparative content and guides |
| Local commerce, Click & Collect | Gemini | Optimize Google My Business + LocalBusiness schema |
| Food, recurring purchases | Mix (Perplexity + Gemini) | Update pricing and availability in real time |
It’s not an exact science. But it’s a starting point that helped 4 of my clients multiply their GEO revenue by 2.3x in 8 weeks.
Don’t put all your eggs in the ChatGPT basket. This LLM is great for awareness and top-of-funnel traffic. But conversion happens elsewhere.
The counterintuitive truth: more traffic = fewer sales?
Almost. If you optimize only for volume, you attract visitors who aren’t ready to buy. And you dilute your overall conversion rate.
One of my electronics clients saw ChatGPT sessions jump from 8,000 to 22,000 in 2 months. Conversion rate dropped from 4.1% to 1.8%. Why? His content was attracting overly broad queries: « difference between SSD and HDD, » not « best 1TB SSD 2025 cheap. »
Conversely, by targeting only purchase-intent queries on Perplexity, traffic dropped 30% but revenue was up 47%.
The goal isn’t to be everywhere. It’s to be where the user pulls out their credit card.
Prioritize your effort, not your LLMs
In summary:
- ChatGPT: volume, inspiration, top-of-funnel traffic. Low conversion rates but broad reach.
- Perplexity: strong purchase intent, high conversion, higher average order. Best for high-engagement products.
- Gemini: local, accessibility, complément. Essential if you have stores.
Don’t be blinded by traffic numbers. Measure conversion by LLM. Segment your reports. And reallocate your effort.
If you don’t know how many sales each LLM generates on your site today, you’re flying blind.
Is your electronics catalog really optimized for Perplexity? Or are you still pouring all your GEO budget into ChatGPT because « everyone talks about it »?
A GEO audit to identify your profitable LLM
I don’t sell you a one-size-fits-all method. During the discovery call, I pull your analytics live, segment your sessions by LLM, and show you where your real sales are. In 30 minutes, you’ll know which chatbot to prioritize for your catalog.
Book a strategic call — 45 minFrequently Asked Questions
Which LLM is best for e-commerce in 2025?
There’s no universal winner. ChatGPT drives discovery traffic, Perplexity converts better on precise purchase intent, Gemini excels in local. The best depends on your catalog: technical? Prioritize Perplexity. Fashion? ChatGPT. Local? Gemini.
How do I measure conversion rate by LLM?
Use UTM parameters on links from each LLM, or segment in Google Analytics 4 by traffic source. For Perplexity, the referrer is ‘perplexity.ai’. For ChatGPT, ‘chatgpt.com’. Gemini currently comes through Google Search (no distinct referrer yet).
Should I abandon ChatGPT for GEO?
No. ChatGPT remains excellent for brand awareness and informational queries. But if your goal is immediate conversion, invest more effort in Perplexity and Gemini. Balance based on your priorities.
How do I optimize my site for Perplexity?
Perplexity indexes your pages and cites them as sources. To get cited, structure pages with schema.org markup (Product, Article), publish visible customer reviews, and keep pricing current. Favor short, factual content in your pages.
Is Gemini relevant for a 100% online e-commerce site?
Gemini is less useful for pure-play sites with no local presence, since it relies heavily on geographic data. However, it can still reference your pages in contextual Google answers. Stay active on Google My Business if you have showrooms or pickup points.
