ChatGPT, Perplexity, Google: I tested an e-commerce site’s visibility. Here are the numbers.
Summarize this article with AI
One morning, Google wasn’t enough. The wake-up call.
A client calls me on a Tuesday morning. An e-commerce site with 4,200 pages, 17,000 organic sessions per month. Nothing massive—healthy traffic, built on solid technical SEO and semantic silos delivered 14 months earlier. But since February 2026, he’s noticed a shift. A 22% drop on his top 5 historical queries. No penalty. No bug. Just… fewer clicks. I open his Search Console. Positions are stable. CTR, though, has tanked. AI Overviews are appearing above his snippets. Users read Google’s summary and don’t scroll down anymore. Then I do something simple. I open ChatGPT, type his 5 main queries. Zero. Not a single mention. On Perplexity, same thing. The site is invisible. Yet its pages are technically flawless. Dense internal linking, SEO entities well-structured. The question shifts. It’s no longer « Is my site well-ranked on Google? » It becomes: Does my site exist anywhere other than Google? I decide to launch a full-scale test. One full month tracking 50 keywords across 3 engines to understand who captures what. This client’s site becomes my lab. What I discovered goes beyond theoretical debates about SEO’s death.50 keywords, 3 engines: the test protocol
You can’t steer what you don’t measure. So we set a framework. Here’s how we worked.
First, we selected 50 strategic keywords for this site. Product queries, catégories, brands, comparisons. Average monthly search volume: 1,400. All tracked in Search Console for over a year. We entered them, one by one, in 3 environments:
- Google: classic SERP (private navigation, no personalization cookies);
- ChatGPT: model with web search enabled, question posed naturally, no context bias;
- Perplexity: same query, « web search » mode, default settings.
For each query, we recorded three things: site presence (yes/no), position if it appeared, and the source listed in the AI response (the URL cited or the database used). We repeated the exercise 3 times over 4 weeks to neutralize variation. 600 total checks.
Raw results, after compilation:
Google classic SERP: 47 queries in top 3.
Google AI Overviews: the site was cited as a source in 22 out of 50 cases.
ChatGPT: only 3 citations.
Perplexity: 7 citations.
The site dominates its market on Google, no question. But it’s nearly invisible in conversational AI responses. ChatGPT doesn’t mention it. Perplexity barely does. And AI Overviews, while Google includes the site, don’t generate enough clicks to offset CTR erosion.
It’s not a content issue. Its pages are rich, structured, with product data, FAQs, reviews. The problem lies elsewhere: AIs don’t build their index like Google does.
Why ChatGPT ignores your site. And it’s not a semantic story.
People often think a strong Google ranking is enough to appear on ChatGPT or Perplexity. It’s an expensive mistake.
These generative AIs use a language model, then activate a search engine for the response (Bing for ChatGPT, in-house index for Perplexity). It’s not a classic crawl. The system picks sources with strong brand authority: frequently cited sites, pages with high social engagement, mentions on Reddit or Wikipedia. An AI’s citation algorithm looks more like a reputation score than a PageRank.
I lived this recently with another client, an 8-person SME. Their site was technically perfect: schema markup, silo-optimized pages down to the keyword. But it was never cited in ChatGPT. Why? Because no other site mentioned it. It spoke about itself. That’s a monologue. ChatGPT and Perplexity reward dialogue between sites.
I asked ChatGPT « What’s the best [product] for [use]? » The answer consistently cited 3 to 4 brands. None were niche sites—they were recognized media, opinion leaders, pages with thousands of backlinks. On-page SEO doesn’t carry weight here. The mindset shifts: you’re no longer optimizing for a crawl, you’re optimizing for a citation.
Result: if your brand doesn’t exist on spaces these AIs scan (forums, media, YouTube, public social networks), you’re invisible. The good news is you can reverse this in weeks. That’s what we did on the test site.
From 3 to 9 citations in ChatGPT: the action plan applied
Faced with initial results, we decided to act fast. No redesign, no hollow new content. We implemented 4 actions, centered on brand SEO and citation.
1. Conversational authority audit.
I identified the 20 most-cited sites by ChatGPT across our 50 keywords. It’s manual work. Every response was scrutinized to list sources. We mapped patterns: 70% of citations came from specialized forums (Reddit, Quora) and media sites. Result: our site was absent from these channels.
2. External mention activation.
We launched 4 Reddit interventions, providing expert answers without spam, with links to product pages where context allowed. At the same time, we placed a guest article on a partner blog. These mentions don’t target direct traffic; they target entry into ChatGPT’s index. Every link is a business card.
3. Entity structuring.
We reworked product entity markup in schema.org to clarify the link between brand, catégories, and products. Nothing new in technical SEO, but coupled with external mentions, it creates a coherence ChatGPT recognizes. The « brand » entity becomes a subject reference.
4. Trust pages.
We created 4 « Why choose us » pages with verifiable data, customer quotes, certifications, third-party reviews. Pages built to be read by an AI, not humans seeking emotion.
Result after 8 weeks: across the 50 keywords, ChatGPT cites our site 9 times. Perplexity, 14 times. The dedicated UTM traffic « source=chatgpt » reached 327 visits in one month. Before, zero.
This traffic is qualitative. Visitors land on the exact page mentioned by the AI, with strong intent. Bounce rate 39%, average session time 3 min 12 s, 2.1 pages viewed. An additional signal for Google, which continues to lift our pages.
Let’s break down the numbers. The site’s organic traffic grew from 17 000 to 17 900 monthly sessions. Here’s what drove that increase.
How AI traffic contributes to overall growth
Decomposition of the 5.3% traffic increase
The real question: does AI cannibalize Google traffic?
This is the fear du jour. « AI Overviews are stealing my clicks. » « ChatGPT is draining my traffic. » I wanted hard numbers.
During the test period, the site’s overall organic traffic grew 5.3%. Classic Google traffic stayed flat. The 17,000 monthly sessions rose to 17,900. This growth is multifactorial: position improvements, expansion into new long-tail queries.
Yet across the 50 tracked keywords, average top-3 CTR dropped 14%. This paradox explains itself: AI Overviews capture clicks on very général queries, but the site compensates with better coverage on specific questions where the AI produces no summary.
The « opportunity cost » calculated on the historical top 5 would be around 1,200 visits per month. But the net gain, including the 327 AI visits and long-tail growth, yields a positive balance. In other words, the pie shifts; it doesn’t shrink.
I’m not saying this holds everywhere. With another client, pure play in home decor, it’s the opposite: traffic dropped 17% in 6 months, with no offset. The difference? Zero presence on AI citation channels. Their site dominated Google but was invisible on ChatGPT. Today, they’re chasing competitors who invested in conversational reputation.
The finding is simple: organic traffic doesn’t die, it mutates. Those who integrate AI as an additional channel capture its value. Others watch their curve slowly erode.
What it costs to do nothing. And what you gain by anticipating.
Take two e-commerce sites, same market, same size. Site A runs a test. Site B, a direct competitor, doesn’t move. In February 2026, their traffic curves are similar.
April 2026: Site A totals 18,227 sessions (Google + AI). Site B stalls at 16,800. The volume difference is modest, but it compounds each month. The 327 AI visits form a flywheel that spins faster. Every week, new queries show Site A in ChatGPT. The phenomenon feeds itself: the more you’re cited, the more likely you’ll be cited again.
Meanwhile, Site B quietly loses ground on AI Overviews. It never appears as a preferred source. Worse, it suffers the perverse effect of competitor citations: when ChatGPT mentions another site, it eclipses theirs for that query. Oblivion comes fast.
The outlook for 2027 is clear. According to internal BMO Academy research, the share of queries routed through AI agents will represent 15 to 20% of web traffic within 18 months. It’s now a distinct acquisition channel. And as with traditional SEO, first movers take the best spots.
I also see a side benefit among those playing the citation game: their brand image strengthens. Appearing in a ChatGPT response is an authority transfer. Users perceive the site as a reference, better than any ad. It’s a free layer of trust.
Does this replace SEO? No. Fundamentals remain the foundation. Without architecture, without technical work, you’ll never hit top 3. Top 3 no longer guarantees click exclusivity. You need to add a layer: AI visibility. 80% of sites still ignore this layer.
Does your site exist on ChatGPT?
47 times in Google top 3. 9 citations in ChatGPT. 14 on Perplexity. 327 additional monthly visits.
This test shows traffic doesn’t vanish—it recomposes. Rules change, but opportunity is real. On one condition: accept that SEO now plays in conversational responses, not just in the SERP. Brand and reputation weigh more than keyword density.
Look at your site. Type your 3 main queries into ChatGPT. Ask it for an opinion, a recommendation, a comparison. Are you cited? If not, you’ve already lost market share without knowing it. No panic—this fixes. Start now, before competitors occupy conversational space alone.
Winning the AI race doesn’t demand massive investment, but activating the right levers with method. The DOSE framework I teach at BMO Academy integrates this citation and brand dimension. This test illustrates it.
So, does your site deserve to be cited by an AI?
Visualize your AI presence in 30 minutes
In a live audit, I show you exactly how your site appears on ChatGPT, Perplexity, and AI Overviews. We identify the 3 priority actions to shift your visibility.
Book a strategic call — 45 minFrequently Asked Questions
Is ChatGPT and Perplexity traffic measurable in Google Analytics?
Yes, if you tag your URLs with precise UTM parameters (source=chatgpt, source=perplexity). ChatGPT doesn’t pass a standard referrer. Without UTM, these visits often classify as direct or are lost. We tracked 327 visits using these tags.
Do I need to optimize my site for ChatGPT with AI-specific content?
No, this isn’t about creating pages for bots. But you must structure your entities (brand, products), earn mentions on sites AIs exploit (Reddit, media, YouTube), and build trust pages. Goal: become a source AIs want to cite.
How long before appearing in ChatGPT?
I saw first citations appear 5 weeks after initial mentions. Timing depends on model training frequency and index updates. Plan for 6 to 8 weeks to see concrete results across 50 keywords.
If I rank well on Google, does that help with AI Overviews?
Partly. For Google AI Overviews, top-3 placement increases your odds of being cited as a source. But it’s not automatic: the algorithm favors source diversity and brand reputation. Some #1 sites miss out if conversational reputation is weak.
Does Perplexity have real impact on e-commerce traffic?
Currently modest, but rising. In my test, Perplexity generated less traffic than ChatGPT (14 citations vs. 9, and less frequent clicks). But the audience is technical and qualified. If you sell information-rich products, this channel is worth it.
