Everyone’s talking about AI shopping features. But do they actually lead to sales?
The research tells conflicting stories. Some studies show AI visitors convert 23 times better than organic search. Others show they convert worse than almost every other channel.
Both claims have data behind them. Here’s what’s actually happening.
The Conversion Question
AI assistants are becoming shopping tools. ChatGPT handles around 50 million shopping queries daily. Perplexity lets users buy without leaving the app.
But traffic and conversions aren’t the same thing. The question merchants care about: when someone finds my product through AI, do they actually buy?
The answer is complicated.
The Optimistic Data
Several studies paint AI traffic favourably:
Ahrefs found AI visitors convert 23x better than organic search.
In their analysis, AI platforms drove just 0.5% of total traffic but accounted for 12.1% of signups. Per visitor, AI sources dramatically outperformed traditional search.
Adobe research shows rapid improvement.
From July 2024 to February 2025, AI referral traffic grew 10x in the United States. More importantly, the conversion gap is closing. In July 2024, AI traffic was 43% less likely to convert than other traffic. By February 2025, that gap had shrunk to just 9%.
Engagement metrics are strong.
Studies consistently show AI referral visitors:
- Stay nearly 10 minutes per session (above average)
- Have 23% lower bounce rates
- View 12% more pages per visit
These aren’t casual browsers. They’re engaged users.
The Pessimistic Data
Other research reaches different conclusions:
E-commerce conversion data tells a different story.
A comprehensive study analysed nearly 1,000 e-commerce sites over 12 months, tracking $20 billion in sales across 160 million sessions. The finding: AI search referrals convert “far worse than Google search traffic.”
When ranked against other channels, AI traffic underperformed affiliate traffic, organic search, email, and direct visits. Only paid social performed worse.
Revenue per session from AI sources was significantly lower than traditional channels.
Why Both Can Be True
The data seems contradictory, but it’s measuring different things:
B2B vs. B2C:
The Ahrefs study measured signups for a B2B SaaS product. High-consideration purchases where research matters. The e-commerce study measured retail purchases—often lower consideration, more impulse-driven.
AI excels at research-heavy decisions. It may underperform for quick purchases.
Signups vs. purchases:
Signing up for a free trial is different from spending money. AI traffic might drive excellent lead generation while underperforming for immediate revenue.
Traffic volume:
AI traffic is still tiny—roughly 0.2% of total e-commerce sessions. Small sample sizes produce volatile conversion rates. One study’s high-converting AI segment might be another’s statistical noise.
Time period:
AI shopping is evolving monthly. Data from July 2024 may not reflect December 2025 behaviour. Early adopters using ChatGPT for shopping are different from mainstream users discovering it now.
How People Actually Use AI for Shopping
The research gap makes more sense when you understand user behaviour.
AI as research tool, not checkout:
Most people use AI to narrow options, not complete purchases. The journey looks like:
- User asks ChatGPT: “What’s the best laptop for video editing under £1,500?”
- ChatGPT recommends three options with reasoning
- User researches those specific products (reads reviews, checks specs)
- User visits retailer directly or searches Google for best price
- User purchases
In this journey, AI shaped the decision but gets zero attribution credit. The sale shows as direct traffic or branded search.
The multi-touchpoint reality:
Purchase decisions rarely happen in one step. AI might be the first touchpoint that narrows consideration from “I need a laptop” to “I want this specific model.”
That’s valuable influence. But it doesn’t show up as AI-driven conversion.
The Attribution Problem
This is the core issue: we can’t accurately measure AI’s impact on purchases.
What we can track:
- Direct referrals from chat.openai.com or perplexity.ai
- In-app purchases through Perplexity’s checkout
- Users who click through and buy in the same session
What we can’t track:
- Users who research on AI, then buy later
- Users who ask AI for recommendations, then Google the product
- Users who see AI suggestions, leave, and return via direct navigation
- Word-of-mouth from AI recommendations (“ChatGPT said this was good”)
Traditional last-click attribution captures a fraction of AI’s influence. The true impact is larger than any referral report shows.
Zero-click answers compound the problem:
AI often answers shopping questions without users clicking anything. “What’s a good gift for a 10-year-old who likes science?” might get a complete answer in the chat. The user never clicks a link, but they know what to buy.
What This Actually Means for E-commerce
Don’t expect direct conversion tracking.
AI-driven sales often won’t appear as AI-driven sales. Accept measurement limitations.
Think of AI as top-of-funnel.
AI shapes consideration and research. It’s an awareness and discovery channel, not a direct response channel. Evaluate it like you’d evaluate PR or brand advertising—valuable but hard to measure precisely.
Watch proxy metrics:
- Branded search volume (is it increasing?)
- Direct traffic trends (are more people arriving knowing what they want?)
- Overall conversion rate (are visitors more qualified generally?)
If AI is working, these metrics should improve even if AI referral traffic stays small.
Focus on being the recommendation.
AI recommends products based on quality, reviews, and fit for user needs. The best AI strategy is also the best business strategy: have a good product that people talk about positively.
Don’t abandon proven channels.
Google Shopping, Meta Ads, email marketing—these have clear attribution and proven ROI. AI is additive, not a replacement. Don’t shift budget from working channels to chase uncertain AI traffic.
The Realistic Outlook
AI shopping will likely matter more over time. Usage is growing 7x year-over-year. The platforms are investing heavily in commerce features. Eventually, AI will be a meaningful traffic source.
But today, it’s a small, hard-to-measure channel with unclear ROI. The research showing great conversion rates and terrible conversion rates can both be accurate—for different contexts, timeframes, and business types.
Monitor it. Ensure your product data is AI-readable. Don’t ignore it. But don’t bet your business on it either.
The merchants who’ll benefit most from AI shopping in 2026 are probably the ones building great products and strong brands in 2025—not the ones chasing AI optimisation tactics.
Research sources: BMON e-commerce study (December 2025), SE Ranking AI traffic research, Adobe Digital Trends analysis. AI shopping features and user behaviour continue to evolve.
