Measuring One Engine Misses 89%
An analysis of 680 million AI citations found only 11 per cent of domains cited by both ChatGPT and Perplexity. An independent study of 118,000 responses produced the same figure; a third measurement gave 12 per cent. Three different methodologies, one conclusion.
The headline stops there. Inside the data sits an unexpected inverse relationship: the engine sending the most traffic draws on the narrowest pool of sources. By one measurement roughly 87 per cent of AI referral traffic comes from ChatGPT, yet ChatGPT runs a web search on only about a third of queries and draws heavily on encyclopedic sources.
Perplexity cites far more sources per answer but accounts for around 3 per cent of referral traffic. The place where you are most easily cited may be the place sending you the fewest visitors.
What the Data Shows
BU BÖLÜMÜN ÖZETİ
- Overlap is very low
- Source preferences differ
- Even one company’s two products diverge
- Brand recommendations disagree too
Four measurements build the picture together.
Overlap is very low
Asked the same 100,000 questions, the two engines share only 11 per cent of cited domains. Around 37 per cent appear only in one, 52 per cent only in the other.
Source preferences differ
Roughly half of one engine’s top citations come from encyclopedic sources; the other’s come from community forums. A third leans on video and multimodal content.
Even one company’s two products diverge
Google’s summary feature and its conversational interface cite the same address only around 14 per cent of the time, even when reaching similar conclusions. “Visibility on Google” is no longer a single thing either.
Brand recommendations disagree too
One comparison measured the engines disagreeing on which brand to recommend about 62 per cent of the time. Cases were reported where citation volume for the same brand differed by hundreds of times across platforms.
What This Figure Means
BU BÖLÜMÜN ÖZETİ
- One engine is a small slice
- Traffic and citation do not gather in the same place
- One strategy does not win three engines
- Mind where the figures come from
Measurement on a single platform does not show the picture.
One engine is a small slice
Tracking only one engine leaves most of the citation landscape outside your field of view. A brand looking strong in one may never appear in another.
Traffic and citation do not gather in the same place
An engine with high citation counts does not necessarily send many visitors. These two indicators need tracking separately; success in one may produce no return in the other.
One strategy does not win three engines
Because source preferences differ structurally, appearing everywhere with the same content format is difficult. Common ground exists — structured, extractable, current content — but the diverging portion is large.
Mind where the figures come from
A significant share of measurement in this field comes from companies selling visibility-tracking tools. The 11 per cent figure being confirmed by three independent methods lends confidence; single-source claims warrant caution.
Whose Problem Is This?
BU BÖLÜMÜN ÖZETİ
- Those sitting comfortably
- Those under pressure
- Those outside it
- How it compounds
Exposure follows where your customers ask their questions.
Those sitting comfortably
Brands known in their field and mentioned across third-party sources. The common denominator across engines is brand recognition and external citation; businesses with both appear in more than one place.
Those under pressure
Businesses measuring visibility on one platform and treating it as overall success. This group never notices the customers it loses on engines where it does not appear.
Those outside it
Businesses whose customers arrive through direct contact and referral are not affected today. The question returns as soon as new customer sources are sought.
How it compounds
Measurement happens on one engine, results look good, the same content format continues, and absence elsewhere goes unnoticed. An unmeasured channel behaves like a channel that does not exist.
What to Do About It
BU BÖLÜMÜN ÖZETİ
- Ask your customers where they ask
- Test your own brand manually
- Invest in the common ground
- Track traffic and citation separately
All four can be done this month and none requires buying a tool.
Ask your customers where they ask
Platform choice belongs on customer behaviour rather than assumption. Asking a few customers where they first heard of you produces a more definite answer than any report.
Test your own brand manually
Put the three questions your customers would ask to different engines and look at the sources returned. The test is free, takes ten minutes and shows directly where you are absent.
Invest in the common ground
Structured content that opens with a direct answer, carries sources and stays current works on every engine. That foundation belongs before any platform-specific tuning.
Track traffic and citation separately
Being widely cited on one engine may bring no visitors. Knowing which channel produces citations and which produces visits changes the budget decision.
The Digital Side
BU BÖLÜMÜN ÖZETİ
- Separate budgets are not required
- Third-party visibility is a shared signal
- Freshness weighs more on some engines
- Budget cannot be split without measurement
The most practical consequence of this divergence shows up in budget allocation.
Separate budgets are not required
Common-ground work serves all engines at once. For the diverging portion, weighting toward the engine that concentrates traffic is a reasonable sequence.
Third-party visibility is a shared signal
However different their source preferences, being mentioned elsewhere carries weight across all of them. Press, sector publications and community platforms matter for that reason.
Freshness weighs more on some engines
One engine was measured giving markedly more weight to recent content. Updating existing pages regularly can do as much work as producing new ones.
Budget cannot be split without measurement
Allocation made without knowing what arrives from where rests on guesswork. It is why we build measurement first in digital consulting.
A Solid Digital Foundation
BU BÖLÜMÜN ÖZETİ
- Crawler access is defined individually
- Access rules need periodic review
- Content should render server-side
- The foundation is built once
Different systems reach your site with different crawlers, and each one’s access is managed separately.
Crawler access is defined individually
Every automated crawler reaching your site arrives with its own identity, and access rules can be written for each separately. Google Search Central explains how crawler access is managed and which rule applies to which crawler. A crawler blocked unintentionally means disappearing entirely from that channel.
Access rules need periodic review
A rule written years ago may not cover systems that exist today. An annual check is enough but necessary.
Content should render server-side
Where a page’s core information is produced only by code running in the browser, some systems never see it. On critical fields like price and specification, that risk is significant.
The foundation is built once
Access rules and content delivery, once organised, apply to every new system as well; we cover the setup within process and e-commerce consulting.
Frequently Asked Questions
Sık Sorulan Sorular
The one your customers use. Traffic share concentrates on one platform, but your particular audience may sit elsewhere.
No. Common ground — structured, sourced, current content — works across all of them. Weighting toward one platform covers the diverging portion.
Their source pools differ structurally. One leans on encyclopedic sources, another on community platforms; the same content behaves differently in each.
The 11 per cent result was confirmed by three separate methods, which is a strong signal. Most measurement in this field nonetheless comes from companies selling tools; single-source claims deserve caution.
The simplest method is manual testing: put your customers’ likely questions to different engines and look at the sources cited. Free and immediate.
It varies by engine. A platform with high citation counts may send few visitors, which is why the two indicators need separate tracking.
Source: Platform comparison based on an analysis of 680 million AI citations; independent verification study covering 118,000 responses; Conductor’s referral traffic measurement and BrightEdge’s brand recommendation comparison. A significant share of measurement in this field is published by companies providing visibility-tracking services.
