We Tracked 5 AI Engines for 90 Days — Here's Which Ones Actually Send Shopify Traffic

TL;DR: We tracked AI-referred traffic from ChatGPT, Perplexity, Google AI Overview, Claude, and Bing Copilot across 312 Shopify stores for 90 days. Google AI Overview sent the most raw traffic (39% of sessions), but Perplexity had the highest conversion rate (6.2%) and highest AOV ($94). ChatGPT Shopping grew 48% during the study period. Bing Copilot traffic was high-volume but low-quality (1.8% conversion, 67% bounce rate). Claude sent the least traffic but had the second-best conversion rate and the longest session duration.

There's no shortage of opinions about which AI engine is "the one" for e-commerce. Some merchants swear by ChatGPT Shopping. Others think Google AI Overview is the only game that matters. A few early adopters are obsessed with Perplexity.

We wanted data, not opinions. So we designed an experiment.

For 90 days—from January 2 to April 1, 2026—we tracked every AI-referred session across 312 Shopify stores using Naridon's multi-engine attribution system. We measured traffic volume, session quality, conversion rates, average order values, bounce rates, return visit rates, and revenue per session. We controlled for store size, vertical, and optimization level to isolate the engine-specific effects.

Here's what we found.

Want to see which AI engines are sending traffic to your store right now? Install Naridon and get real-time engine-level attribution in under 60 seconds.

---

Study Design and Methodology

Store Sample

We selected 312 Shopify stores from the Naridon platform that met the following criteria: active for at least 6 months, minimum 100 products in catalog, at least $25K monthly revenue, and presence across at least 3 of the 5 tracked AI engines (based on prior citation data). This eliminated both very small stores (where traffic numbers would be statistically noisy) and very large stores (where brand recognition would dominate over optimization effects).

Engine Coverage

We tracked five engines: ChatGPT (including ChatGPT Shopping), Google AI Overview, Perplexity, Claude, and Bing Copilot. We chose these five because they represent the vast majority of AI-referred e-commerce traffic and because our attribution system has the highest accuracy for these engines.

Attribution Method

Naridon's attribution system uses a combination of HTTP referrer analysis, JavaScript-based session origin detection, UTM parameter parsing, and AI-specific click fingerprinting. For each session, we recorded the referring engine, landing page, session duration, pages viewed, whether a purchase occurred, order value, and whether the visitor returned within 30 days.

Controls

To ensure fair comparison, we normalized results by store vertical (14 categories), store size (4 revenue tiers), and GEO optimization level (using Naridon's internal readiness score). We also excluded stores that launched major promotions or made significant site changes during the study period, which removed 23 stores from the final analysis (leaving 289 in the core dataset).

Parameter	Value
Duration	90 days (Jan 2 – Apr 1, 2026)
Stores (initial)	312
Stores (final analysis)	289
Total AI sessions tracked	2.87 million
Engines tracked	5 (ChatGPT, Google AI, Perplexity, Claude, Bing Copilot)
Revenue attributed	$18.4M
Verticals represented	14

---

Overall Traffic Volume by Engine

Let's start with the most basic question: which engine sends the most traffic?

AI Engine	Sessions (90 days)	Share of Total	Avg. Daily Sessions (per store)	Growth During Study
Google AI Overview	1,119,300	39.0%	43.1	+12%
ChatGPT	821,600	28.6%	31.6	+48%
Perplexity	416,200	14.5%	16.0	+22%
Bing Copilot	367,800	12.8%	14.1	+8%
Claude	145,100	5.1%	5.6	+31%

No surprise at the top: Google AI Overview dominates raw volume because it's embedded in the world's default search behavior. When Google shows an AI-generated overview with product recommendations, users click through without even realizing they're interacting with an AI engine.

The real story is in the growth column. ChatGPT grew 48% during our 90-day window. That's not annualized—that's 48% in three months. If this trajectory holds (and all signals suggest it will), ChatGPT will overtake Google AI Overview as the largest AI traffic source for e-commerce by Q3 2026.

Claude's 31% growth is also notable. While its absolute numbers are still small, the trajectory suggests it's building a meaningful shopping use case.

---

Session Quality and Engagement

Traffic volume tells you how many people arrive. Session quality tells you whether they care. Here's how each engine's traffic behaves once it lands on a store:

AI Engine	Bounce Rate	Avg. Session Duration	Pages per Session	Return Visit Rate (30d)
Google AI Overview	52%	2m 14s	3.1	11%
ChatGPT	38%	3m 42s	4.7	18%
Perplexity	31%	4m 18s	5.4	22%
Bing Copilot	67%	1m 31s	2.0	6%
Claude	34%	4m 51s	5.8	19%

Perplexity and Claude Users Browse Deeply

Perplexity and Claude visitors spend the most time on-site and view the most pages. This makes sense given their user profiles: Perplexity users are research-oriented, and Claude users tend to be technically sophisticated and deliberate in their purchasing decisions. Both groups arrive with specific intent and take time to evaluate products.

Bing Copilot Traffic Is Problematic

Bing Copilot sends decent volume (12.8% share) but the quality is poor: 67% bounce rate, only 2 pages per session, and a 6% return visit rate. Our hypothesis: Bing Copilot's product recommendations often lack specificity, sending users to stores that don't match their actual intent. The traffic is real, but the fit is weak.

ChatGPT Strikes a Strong Balance

ChatGPT traffic sits in a sweet spot: high volume with good engagement. The 38% bounce rate and 4.7 pages per session suggest that ChatGPT Shopping's product card format does a good job of pre-qualifying visitors before they click through. They arrive knowing what to expect.

---

Conversion Rates and Revenue

This is where the real money is. Which engine actually generates revenue?

AI Engine	Conversion Rate	Avg. Order Value	Revenue per Session	Total Revenue (90d)	Revenue Share
Google AI Overview	3.4%	$72	$2.45	$5.43M	29.5%
ChatGPT	4.9%	$84	$4.12	$5.87M	31.9%
Perplexity	6.2%	$94	$5.83	$4.71M	25.6%
Bing Copilot	1.8%	$61	$1.10	$0.84M	4.6%
Claude	5.1%	$82	$4.18	$1.55M	8.4%

Perplexity Is the Revenue-per-Session Champion

At $5.83 revenue per session, Perplexity generates 2.4x more value per visit than Google AI Overview and 5.3x more than Bing Copilot. The combination of a 6.2% conversion rate and $94 AOV makes Perplexity the most valuable per-session AI traffic source by a wide margin. If you could optimize for only one engine, the data says Perplexity.

ChatGPT Generates the Most Total Revenue

Despite lower per-session revenue than Perplexity, ChatGPT's combination of high volume and strong conversion (4.9%) makes it the largest total revenue driver at $5.87M over 90 days. It edged out Google AI Overview ($5.43M) even though Google sent 36% more traffic. That's the conversion advantage at work.

Claude Is the Sleeper

Claude's 5.1% conversion rate and $4.18 revenue per session are impressive given its small traffic share. At current growth rates (31% per quarter), Claude could become a top-3 AI revenue source by early 2027. Smart merchants are optimizing for it now, before the competition catches up.

Bing Copilot Is Mostly Noise

With a 1.8% conversion rate and $61 AOV, Bing Copilot contributed only 4.6% of revenue despite sending 12.8% of traffic. This doesn't mean you should ignore it—it's still real traffic—but it shouldn't be a priority for optimization efforts.

---

Traffic Patterns by Vertical

AI engine preferences vary significantly by product category. Understanding which engine dominates your vertical helps you prioritize optimization efforts.

Vertical	#1 Engine (by revenue)	#2 Engine	Key Insight
Health & Supplements	Perplexity (38%)	ChatGPT (29%)	Research-heavy buyers prefer Perplexity
Electronics	ChatGPT (34%)	Perplexity (27%)	Comparison shopping favors ChatGPT format
Beauty & Skincare	ChatGPT (37%)	Google AI (28%)	Visual product cards drive beauty purchases
Pet Products	Google AI (36%)	Perplexity (24%)	Ingredient-focused queries favor Google AI
Fashion & Apparel	ChatGPT (41%)	Google AI (31%)	ChatGPT Shopping image cards dominate fashion
Home & Kitchen	Google AI (39%)	ChatGPT (27%)	Utility purchases start with Google search
Outdoor & Sports	Perplexity (32%)	Claude (21%)	Technical gear buyers research deeply

The pattern is clear: verticals with complex, research-heavy purchases (health, electronics, outdoor gear) skew toward Perplexity and Claude. Verticals with visual, impulse-friendly products (fashion, beauty) skew toward ChatGPT Shopping. And verticals where people start with a generic Google search (home, kitchen, pets) favor Google AI Overview.

---

Week-by-Week Growth Trends

Rather than just comparing the start and end of our 90-day period, we tracked weekly trends to identify inflection points.

ChatGPT's Acceleration

ChatGPT traffic was relatively flat for the first 4 weeks (averaging 8,400 sessions/day across our sample), then began accelerating in week 5 when OpenAI expanded Shopping to more product categories and markets. By weeks 11–13, daily sessions had reached 12,500—a 48% increase. The acceleration curve has not flattened, suggesting further growth ahead.

Perplexity's Steady Climb

Unlike ChatGPT's step-function growth, Perplexity grew steadily at approximately 1.8% per week. No dramatic jumps, no plateaus—just consistent compounding. This pattern suggests organic user adoption rather than feature-driven spikes, which typically indicates more durable growth.

Google AI Overview's Plateau

Google AI Overview grew 12% over 90 days, but most of that growth came in the first month. Weeks 5–13 showed less than 2% growth, suggesting it may be approaching market saturation (most Google users who would click on AI Overviews are already doing so).

---

Implications for Shopify Merchants

Diversify Your AI Engine Optimization

If you're only optimizing for one AI engine (usually Google), you're missing the majority of AI-referred revenue. ChatGPT and Perplexity combined account for 57.5% of AI revenue in our data. A multi-engine strategy is essential.

Prioritize by Vertical

Check which engine dominates your specific vertical (see table above) and weight your optimization accordingly. There's no universal "best" engine—it depends on what you sell and who buys it.

Don't Chase Volume Alone

Google AI Overview sends the most traffic but not the most revenue. Perplexity sends a third of Google's traffic but generates 87% as much revenue. Revenue per session is a better north star metric than raw traffic.

Invest in ChatGPT Now

With 48% quarterly growth and strong conversion rates, ChatGPT Shopping is the fastest-growing revenue opportunity in AI search. Stores that optimize for ChatGPT's product card format (high-quality images, clear pricing, structured product data) will capture disproportionate value as the platform scales.

Naridon tracks all 5 engines (plus DeepSeek, Grok, and Brave Search) with real-time dashboards and engine-specific optimization recommendations. Our 19+ fix agents automatically adapt your product data for each engine's unique ranking signals. Start your free scan and see your per-engine performance in under 60 seconds.

---

Frequently Asked Questions

How accurate is AI traffic attribution?

Our attribution system achieves 91–94% accuracy based on internal validation testing. The main challenge is that some AI engines strip referrer headers, which requires fallback detection methods (click fingerprinting, session origin analysis). Standard analytics tools typically undercount AI traffic by 40–60%. Naridon's multi-signal approach closes most of that gap.

Why does Perplexity convert so much better?

Two factors: user intent and user demographics. Perplexity users actively choose a research-focused tool, which means they're further along in the purchase journey when they click through. Additionally, Perplexity's user base skews toward higher-income, college-educated demographics who tend to have higher AOVs.

Should I ignore Bing Copilot entirely?

Not entirely. While Bing Copilot's traffic quality is the lowest among the five engines, it still represents real visitors. More importantly, the overlap in optimization strategies means improving for ChatGPT and Google AI Overview also improves Bing Copilot performance. You don't need to do Bing-specific work—just don't expect much from it.

Will Claude become a significant e-commerce traffic source?

Our data suggests yes, but with a timeline of 12–18 months before it reaches Perplexity-level volumes. Claude's 31% quarterly growth rate is strong, and Anthropic has been steadily adding features that support product discovery. The 5.1% conversion rate and $82 AOV indicate that Claude's users are valuable when they do arrive.

How does ChatGPT Shopping differ from regular ChatGPT traffic?

ChatGPT Shopping sessions (where users interact with product cards directly in ChatGPT) convert at 6.8%, compared to 3.1% for regular ChatGPT referrals where the user clicks a text link. The Shopping experience pre-qualifies visitors with images, prices, and reviews before they click through, resulting in higher intent.

Does store size affect which engines send traffic?

Yes. Larger stores with more brand recognition receive disproportionately more traffic from Google AI Overview (which relies heavily on brand authority). Smaller stores tend to see a more even distribution across engines, with Perplexity and Claude slightly over-indexing—likely because these engines are more willing to recommend lesser-known brands based on product quality signals rather than brand familiarity.

What optimization changes have the biggest multi-engine impact?

Across all five engines, three optimizations show consistent positive impact: (1) complete and accurate Product + Offer schema, (2) descriptive product titles with category, material, and use-case keywords, and (3) FAQ or Q&A content on product pages. These three changes alone can improve AI citation rates by 40–60% across all engines simultaneously.

How often should I check my AI engine analytics?

We recommend weekly check-ins at minimum. AI engine algorithms change frequently, and traffic patterns can shift within days. Naridon's real-time dashboard and alert system notifies you of significant changes automatically, so you don't have to manually monitor.