How the Optimly AI Brand Index measures AI retrieval
The weekly panel shows what automated AI agents accessed across the Optimly Index corpus. It is designed to help marketers distinguish active AI use from crawling and training—and turn observable topic signals into better AEO and GEO research.
Why we built it
Marketers need evidence of access before they infer influence
AI systems increasingly retrieve brand information while answering questions, indexing the web, or collecting material for model development. Most marketing dashboards combine those activities into one large “AI traffic” number.
That makes it easy to mistake training collection for buyer interest—or to treat a page request as proof that a model cited or recommended a brand. We built the panel to keep those claims separate.
The result is an observable layer between a brand's published information and an AI-generated answer: which Index pages agents requested, why the request most likely occurred, and how that pattern changed over time.
What the corpus contains
A structured map of brands and the questions around them
The panel covers published Index pages that can be resolved into meaningful marketing entities.
Brand profiles
Canonical pages for individual brands, including the profile a brand can claim and maintain.
Category pages
Market groupings used to understand which categories and competitive sets attract agent attention.
Buyer-topic pages
Pages organized around problem, solution, evaluation, comparison, and overview topics.
Supporting Index pages
Directory and structured record pages that help agents discover and resolve the corpus.
Measurement flow
From a page request to a qualified weekly signal
Record the request
Cloudflare request data records that an automated agent asked optimly.ai for a page during the reporting window.
Identify the agent
Known user-agent and behavioral signals are mapped to an agent, provider, and likely purpose when the evidence supports it.
Resolve the page
The requested path is matched to a published brand profile, category, buyer-topic page, or another Index page type.
Aggregate the panel
Requests are grouped into non-overlapping purpose groups and summarized by page type, brand, category, and buyer topic.
Qualify the comparison
The report checks sample length, matched-page coverage, traffic volume, brand breadth, errors, and overlap before describing a trend.
Agent classification
Not every automated visit means the same thing
Each visit is placed in one non-overlapping group. The groups add up to the full automated panel for the week.
People using AI products
Requests most closely associated with an active AI-product session. This is the panel's closest available proxy for human-directed activity, but it is not a count of people or prompts.
Search engine crawlers
Automated discovery and indexing traffic used by search products. It shows that a page was crawled, not that it appeared in an answer.
AI training bots
Agents that collect page text for model development or future training. No person should be assumed to be waiting for these requests.
Known company, purpose unclear
Traffic that can be attributed to a company but cannot be assigned confidently to active use, search, or training.
Other automated traffic
Automated requests for which the company or purpose is not named clearly enough to support a stronger classification.
For content teams
Why the raw buyer-topic path is still useful
A raw path preserves two useful pieces of evidence: the topic represented by the page and its place in the buying journey. It should be treated as a content-research lead—not as a captured prompt.
/intent/problem/fragmented-brand-data
Explain the problem, symptoms, causes, stakes, and available approaches.
/intent/solution/sovereign-cloud-providers-europe
Define the solution category, use cases, audience, evidence, and selection criteria.
/intent/evaluation/alternative-to-example
Publish alternatives, evaluation criteria, proof, limitations, and decision-stage FAQs.
/intent/comparison/brand-a-vs-brand-b
Create a factual comparison with differences, trade-offs, best-fit scenarios, and sources.
Interpretation rule: validate a path-derived opportunity with first-party customer questions, sales conversations, search demand, competitive coverage, and subject-matter expertise before publishing. The panel tells you where an agent accessed the Index; it cannot tell you the exact user wording or whether the model used the page.
The panel can show
- Which published Index pages an automated agent requested
- Which identified agent or provider was most active
- How traffic was distributed across likely purposes
- Which brands, categories, and buyer-topic pages changed
- How complete and comparable the reporting window was
The panel cannot prove
- The exact prompt or words a person entered
- That the page appeared as a citation or source
- That a model stored, learned, or remembered the content
- That an AI answer mentioned or recommended the brand
- That automated traffic represents market share, demand, or revenue
Frequently asked questions
Definitions for reading and citing the panel
What does the Optimly AI Brand Index measure?
The panel measures automated requests to pages in the Optimly AI Brand Index. It summarizes which agents accessed brand profiles, category pages, buyer-topic pages, and related Index resources during each reporting period.
Does an AI retrieval mean a model cited or recommended a brand?
No. A retrieval proves that an identified automated agent requested a page. It does not prove that the page was cited, quoted, stored, learned, or used to recommend a brand.
Can the panel see the exact question a person asked an AI product?
Usually not. Buyer-topic labels are derived from the address of the Index page an agent requested. They describe the topic and buying stage of that page, but they are not transcripts or exact user prompts.
How can marketers use raw buyer-topic paths?
A raw path can reveal the topic and buying stage that an Index page represents, such as problem, solution, evaluation, or comparison. Marketers can use it as a content-research lead, then validate the opportunity with customer, search, sales, and competitive evidence.
Is the AI Brand Index a measure of total AI market share?
No. It is a panel of activity observed on Optimly's own Index corpus. It can show movement and concentration within that panel, but it is not a census of all AI activity or consumer demand.
How often is the panel updated?
The current report is designed to publish weekly. Every edition identifies its reporting window and discloses when the prior comparison is incomplete, overlapping, or otherwise directional.
See the methodology in practice