How visible is our brand in AI search, and how much actual website traffic does it generate? These two questions need to be examined together today. A review of recent research reveals that there are many more nuances to check between the sources cited in answers and the pathways leading to visits.
We have selected three key signals from the analyses published on October 28 by Muck Rack, Ahrefs, and First Page Sage. All of these are observational studies by external parties and do not represent official updates from search platforms. Rather than taking the results directly as our brand performance standards, we should use them to determine what to measure and which data points to audit.
Signal 1 We Must Also Examine External Descriptions of Our Brand#
Muck Rack’s analysis showed that over 84% of sources cited by AI were classified as earned media and third-party content. In addition to press articles, this includes government and non-profit organization materials, academic resources, encyclopedias, and content from other enterprises. Interpreting this figure solely as the share of news articles would cause you to miss the actual scope of the study. Due to limitations in the published samples and classification descriptions, this ratio cannot be viewed as a universal baseline for all AI searches. Muck Rack Research
From a brand perspective, the crucial takeaway is what kind of narrative is being built outside our own website. Even if your product use cases and application scope are clearly explained on your proprietary pages, you must also review whether conflicting information persists in external materials. However, this study alone does not prove that featuring in a specific media outlet increases citations.
First, select a few questions that customers typically ask when comparing products or considering adoption. By reading the sources appearing in AI answers to those queries, you can check whether the description of your brand is accurate and whether any necessary information is missing. Rather than stopping at the fact that a competitor was mentioned, identifying which materials supported that description will give you more specific targets for revision.
For example, if your service support scope is displayed differently in your company introduction versus an external directory, you can correct that discrepancy first. For text you cannot edit directly, it is better to document the factual corrections needed and their rationale. While this audit does not guarantee citations, it helps mitigate the problem of customers encountering conflicting descriptions.
Signal 2 AI Traffic Metrics Vary Depending on the Sites Surveyed#
Although both reports present the AI traffic share for September 2026, their target audiences and definitions of “AI traffic” differ. Ahrefs measured visits from AI Assistants using a fixed panel, whereas First Page Sage aggregated AI platform sessions across client samples, including Google AI Overviews and AI Mode. Placing these two figures side by side may make them look like they measure the same thing, so let’s examine the sample sizes, periods, and aggregation scopes together.
September 2026 AI Traffic Share and Target Audience#
| Research | Reported Share | Survey Panel & Period | Aggregation Scope |
|---|---|---|---|
| Ahrefs | 0.43% (Based on visits) | Fixed panel of 115,254 sites, June 2025–September 2026 | Visits from AI Assistants. Share of total visits within the panel, not total market share |
| First Page Sage | 6.9% (Based on sessions) | 218 client sites across 12 industries, January 2023–September 2026 | AI platform sessions, including AI Overviews, AI Mode, and direct AI assistant traffic |
Because these two percentages differ not only in sample size but also in aggregation units and the definition of AI traffic, they cannot be directly compared. It is more accurate to track trends independently within each report using consistent criteria.
Source: Ahrefs Benchmark · First Page Sage Report
Furthermore, these findings alone do not prove that AI has cannibalized existing search traffic. Multiple changes can occur simultaneously during the same period, and traffic share also fluctuates based on visit volume from other channels. Before directly translating external study ratios into projected losses or growth for our site, we must examine any changes identifiable in our own first-party data.
Large discrepancies in numbers naturally prompt the question of which side is correct. However, given the differences in survey targets and aggregation methodologies, the gap between these two figures cannot be interpreted as a market growth rate. It is best to first verify the types of visits your site receives before determining how heavily to rely on external benchmarks.
In particular, brand visibility in AI answers and actual user visits to the site need to be tracked separately. Users may obtain sufficient information directly within the answer without visiting, or they may use source links to investigate further. We must remember that these traffic studies specifically examine visits that successfully arrived at the site.
In your monthly audits, prioritize recording brand mentions and source citations for key queries, while separately analyzing AI-driven traffic identified in analytics tools and user behavior post-visit. Meaningful actions such as consultation requests or resource downloads should also be tracked. This prevents us from dismissing the value of visibility simply because traffic is low, or assuming business performance has improved solely because citations increased.
When evaluating post-visit behavior, consider the role of the page. Applying the same conversion criteria to a visit intended for usage verification versus one for evaluating a purchase target can lead to misinterpretation. In stages where visit volume is insufficient, observing which queries and pages generate traffic is more productive than overinterpreting minor numerical fluctuations.
Signal 3 It Is Difficult to Judge Status Based on a Single Answer#
Muck Rack also examined the consistency of cited domains when queries were repeated on the same day. The chart below shows the results of that experiment. A high consistency rate does not imply that the answer is more accurate or that brand performance is stronger.
| Service | Citation Domain Consistency Rate |
|---|---|
| ChatGPT | 41% |
| Gemini | 45% |
| Perplexity | 94% |
- Measurement Conditions: Experiments within research repeating queries on the same day.
- Publication Date: October 28, 2026.
These are observational results under specific research conditions and do not indicate service quality rankings.
Source: Muck Rack Research , October 28, 2026.
Nevertheless, the findings offer useful lessons for auditing your brand. Appearing in a query run once today does not mean your exposure is stable. Conversely, failing to appear once does not necessitate overhauling your entire content strategy. Repeated observations allow you to track which explanations and sources consistently recur.
Start by defining a few questions related to actual customer comparisons and evaluations, and log the service, query wording, and check dates together. Separating recurring queries from newly explored queries makes it easier to trace what changed if and when the AI’s answers shift. While you should record modified pages and dates, avoid immediately concluding that subsequent shifts in answers were caused solely by those edits.
Observation logs should capture not only whether the brand name appeared, but also the context of the mention and the specific linked source page. If the same erroneous description appears repeatedly, you can prioritize auditing the underlying source material. The goal is to use logs to pinpoint actionable problems rather than focusing on a screen that resembles a transient ranking.
What Should We Use to Measure Our Brand’s AI Search Performance?#
As we review these findings together, the first practice we should update is our AI search reporting methodology. Separating how external sources describe us from how actual traffic is generated, while logging observation dates and conditions, will make next month’s comparisons much clearer. Before judging success or failure based on a single number, we must also verify the scope and context of that result.
Starting small is completely acceptable. Select a few critical customer questions, log their answers and sources, and examine post-visit behaviors driven by AI traffic in your existing analytics tools. You can then systematically fix errors or gaps in information discovered through this repetitive process. This ensures that when you read the next piece of research, you can evaluate it in connection with your own recorded data.
Questions Raised by This Signal#
If a customer encounters our brand in an AI answer but does not visit our site, how can we evaluate the value of that touchpoint?
About Search Signals#
Search Signals is a series that observes changes in the SEO/GEO landscape based on official updates from major search platforms such as Google Search Console and NAVER Search Advisor, interpreting their implications. Rather than focusing solely on new features, we concentrate on the direction pointed by these changes and the critical questions practitioners must re-examine, drawing on reports from trusted services and institutions.
Published on the 15th and last day of every month.
