SpyderBot Documentation
AI Search and global exposure
AI Search has two different measurement opportunities: a search can produce an AI answer, and that answer can include a brand. SpyderBot therefore separates exposure across search opportunities from performance within generated AI answers.
A verified search result with no AI answer is a real observed outcome. A CAPTCHA, failed load, interrupted collection or unresolved answer state is not a verified absence. Search system, surface, locale, language, time and question scope are part of the result.
On this page
Google AI Overviews · Google AI Mode · Bing Search · AI Answer Availability · Search Brand Presence Exposure · Brand Presence within AI Answers · Search Recommendation Exposure and Answer Recommendation Rate · Search Citation Exposure and Answer Citation Presence · Global Organic Visibility Exposure · Global Organic Recommendation Exposure · Global Organic Share of Voice · Observed Search Fan-Out
Google AI Overviews
Google AI Overviews is Google's name for AI-generated summaries presented within Google Search results. An individual summary is an AI Overview. It brings together key information and links that help users investigate a question further. Google determines when to display this feature; it does not appear for every query. Google Search Help ↗
What it means in a SpyderBot report
An observation attributed to Google AI Overviews concerns that specific AI answer and its associated sources. A brand appearing only in an ordinary result elsewhere on the search page is not, by itself, a mention in the AI Overview. Likewise, an ordinary search-result link does not become an AI citation simply because it shares the page with an AI answer.
Read three different outcomes separately:
- Answer availability: Was a usable AI Overview observed for the eligible search opportunity?
- Brand presence or recommendation: Did the AI answer substantively mention or recommend the brand?
- Citation presence: Did the answer include a qualifying reference to the selected domain or source category?
A source can be cited without its publisher being recommended. A brand can also be mentioned without a citation to its own website. These outcomes answer different questions.
If a completed, applicable observation confirms that no AI Overview appeared, it is a verified no-answer outcome for that surface. A blocked or incomplete observation is missing evidence, not proof that the feature or brand was absent.
Example: Your website appears in a regular Google result, but the AI Overview mentions only competitors. That observation does not establish AI Overview brand presence for your brand.
AI Answer Availability ↗ · Brand exposure ↗ · Citation measurements ↗
Google AI Mode
Google AI Mode is a dedicated AI-powered experience within Google Search for exploring questions through generated answers, supporting web links and follow-up questions. It supports more involved exploration and comparisons. Google explains that AI Mode can divide a question into subtopics and search across them. It can also return web links instead of an AI answer when an AI response is not considered sufficiently helpful or reliable. Google Search Help ↗
AI Mode is distinct from the AI Overview shown within a standard results page. Google states that the two experiences may use different models and techniques, so their answers and links can differ. The surface name therefore does not establish a fixed model version. Google Search Central ↗
What it means in a SpyderBot report
Read Google AI Mode results as observations of AI Mode itself. A result from the Gemini assistant or from Google AI Overviews cannot substitute for an AI Mode observation merely because the services share a provider or related model technology.
Conversation context matters. An answer to a fresh question and an answer after several follow-ups are different measurement contexts. Compare results using the declared question, language, location, collection time and session context. A provider's ability to support follow-ups does not mean that a report measured an entire conversation.
Do not interpret a source list as a complete record of the searches performed internally. A description of query fan-out explains possible provider behavior; it does not establish that SpyderBot observed those hidden subqueries.
Example: A fresh comparison question and a follow-up asking specifically about your brand have different eligibility. The follow-up cannot be treated as an independent organic-discovery answer simply because it appears in AI Mode.
Presence within AI answers ↗ · Recommendation measurements ↗ · Observed Search Fan-Out ↗
Bing Search
Bing Search is Microsoft's web search service. It includes ranked web results and other search experiences, some of which provide AI-generated answers. The broad name identifies a search product, not one language model or one uniform AI-answer format. How Bing delivers search results ↗
AI experiences within Bing
Microsoft's current Bing product guide describes AI Overview, a summary within search results, and AI Mode, a conversational search experience. It describes these as the evolution of Copilot Search in Bing, powered by Copilot technology. To distinguish them from Google's similarly named experiences, this documentation uses the qualified labels Bing AI Overview and Bing AI Mode. Availability and behavior can vary by market, device and browser. Microsoft Bing product guide ↗
Older references may use Bing generative search or Copilot Search in Bing. Keep historical results tied to the experience actually observed. A change in branding alone is not evidence that historical and current measurements are equivalent. Bing generative search announcement ↗ · Copilot Search in Bing announcement ↗
What it means in a SpyderBot report
When a report uses the broad label Bing Search, inspect the reported scope and available evidence to understand which experience was collected. That label alone does not establish coverage of every Bing AI experience. Do not infer a specific subtype when the observation does not identify it.
- A normal Bing listing or snippet is not automatically a mention or citation in a generated AI answer.
- A generated answer and its qualifying references are evaluated within the identified answer surface.
- A Copilot assistant response is not automatically a Bing Search observation, even when it uses web search.
- Microsoft Copilot Search for work and school is a separate enterprise search experience; it should not be confused with public Bing web search. Microsoft Learn ↗
Example: A brand appears in a Bing web listing but not in the observed AI answer. That listing alone does not establish AI-answer brand presence. If the record does not identify which Bing AI surface was observed, it cannot support a comparison specifically attributed to Bing AI Mode.
Search exposure ↗ · Presence within AI answers ↗ · Evidence and scope ↗
AI Answer Availability
Unit: Percent of verified search opportunities
How often the observed search opportunities produce a usable AI-generated answer.
How it is calculated
Divide verified opportunities with a generated usable AI answer by all verified applicable search opportunities, then aggregate within the declared balanced scope. Verified no-answer results remain in the denominator; technical unknowns remain missing.
How to interpret it
Availability describes the observed search surface, not availability for all real-world searches. It is separate from brand performance within the generated answers.
What to do next
Check availability before interpreting a low exposure rate.
Illustrative example
Of 100 verified search opportunities, 40 produce an AI answer: availability is 40%.
Related concepts: Search Brand Presence Exposure ↗ · Brand Presence within AI Answers ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Search Brand Presence Exposure
Unit: Percent of verified search opportunities
How often a search opportunity produces an AI answer that substantively mentions the brand.
How it is calculated
Use all verified applicable search opportunities as the denominator. A generated answer with a target mention is positive; a generated answer without a mention and a verified no-answer result contribute no brand exposure. Apply the declared balanced aggregation.
How to interpret it
This includes the effect of whether an AI answer appears. It must not be labelled as brand presence within AI answers.
What to do next
Separate answer availability from the brand’s presence when answers do appear.
Illustrative example
If 40 of 100 verified opportunities generate an answer and 10 mention the brand, exposure is 10%; presence within answers is 25%.
Related concepts: Brand Presence within AI Answers ↗ · Global Organic Visibility Exposure ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Brand Presence within AI Answers
Unit: Percent of generated usable AI answers
How often generated usable search AI answers substantively mention the brand.
How it is calculated
Condition on verified opportunities that generated a usable AI answer. Calculate target mention presence within those answers and apply conditional balanced aggregation. Verified no-answer opportunities do not belong in this denominator.
How to interpret it
With no generated AI answers, this is not applicable, even when search exposure is a valid 0%. It is not the same percentage as exposure across searches.
What to do next
Review the displayed denominator before comparing this value with another platform.
Related concepts: Search Brand Presence Exposure ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Search Recommendation Exposure and Answer Recommendation Rate
Unit: Percent, with two distinct denominators
How often the target is recommended across search opportunities, and how often it is recommended within generated AI answers.
How it is calculated
Exposure uses every verified applicable search opportunity; verified no-answer contributes zero exposure. Answer Recommendation Rate uses only generated usable AI answers in the eligible recommendation task. The same substantive recommendation rule applies, followed by the correct balanced or conditional aggregation.
How to interpret it
Naming, listing or citing a brand does not automatically mean recommending it. The two percentages must have separate labels, values and denominators.
What to do next
Investigate whether the constraint is answer availability or recommendation within available answers.
Related concepts: Organic Recommendation Rate ↗ · AI Answer Availability ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Search Citation Exposure and Answer Citation Presence
Unit: Percent, with explicit scope
How often search opportunities expose the selected citation category, or generated AI answers contain qualifying citations.
How it is calculated
For exposure, divide opportunities with the selected citation outcome by verified applicable search opportunities. For presence within answers, condition on generated usable AI answers. An owned, competitor or third-party category uses its declared domain mapping; multiple categories can coexist in one answer.
How to interpret it
Do not add overlapping category percentages to make a 100% pie. These measurements do not establish visits to cited sites.
What to do next
Open the appropriate source evidence and verify which denominator is displayed.
Related concepts: Owned Citation Presence ↗ · Citation Presence ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Global Organic Visibility Exposure
Unit: Percent of the declared cross-channel opportunity profile
A balanced summary of organic brand exposure across the included assistant and search channels.
How it is calculated
First calculate each channel on its proper opportunity basis: assistant response opportunities and search opportunities, including verified search no-answer outcomes. Combine the channel results using the named comparison profile. The balanced cross-channel profile assigns 50% to assistants and 50% to search; an assistant-only profile is explicitly labelled assistant-only. All planned channel requirements must be met before publishing a full-profile headline.
How to interpret it
The profile is a standardised measurement mix, not an estimate of actual AI-user traffic share. This new exposure metric is not interchangeable with legacy assistant response visibility.
What to do next
Inspect channel breakdowns before deciding where to act.
Related concepts: Organic Visibility ↗ · Search Brand Presence Exposure ↗ · Global Organic Recommendation Exposure ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Global Organic Recommendation Exposure
Unit: Percent of the declared cross-channel opportunity profile
The cross-channel opportunity share on which the target receives an organic recommendation.
How it is calculated
Combine assistant recommendation exposure and search recommendation exposure using the declared channel profile, after each has been aggregated on compatible scope. For the balanced profile, the two channel results receive equal weight. Do not multiply independently aggregated headline availability and conditional recommendation percentages to reconstruct search exposure.
How to interpret it
Channel exposure is calculated from the underlying matched observations. A blended value is not real-world market share or a causal measure of conversion.
What to do next
Read the channel and task breakdowns.
Illustrative example
In a single matched-scope illustration, assistant recommendation is 60%; search generates answers on 50% of opportunities and recommends the brand in 40% of those answers, giving 20% search exposure. The balanced global exposure is (60% + 20%) / 2 = 40%. This shortcut is not valid for independently aggregated multi-scope headline percentages.
Measurement scope and aggregation ↗ · Evidence and availability ↗
Global Organic Share of Voice
Unit: Percent of tracked cross-channel exposure
The target’s share of balanced global brand exposure within the declared comparison set.
How it is calculated
Calculate each tracked entity’s global exposure on the same complete opportunity profile, then divide target exposure by the total tracked-entity exposure. If the total is zero, share is unavailable.
How to interpret it
Changing the brand roster or channel profile changes the comparison basis. The result is not a share of all brands or AI users.
What to do next
Compare like-for-like rosters and profiles.
Related concepts: AI Share of Voice ↗ · Global Organic Visibility Exposure ↗
Measurement scope and aggregation ↗ · Evidence and availability ↗
Observed Search Fan-Out
Unit: Exposed trace or unavailable
Search subqueries or retrieval steps actually exposed by the observed system trace.
How it is assessed
This result contains only the search or retrieval trace actually exposed by the provider or supported observation surface. Without that evidence, fan-out is unavailable. Answer text, citation lists and generic model explanations do not establish hidden subqueries.
How to interpret it
A partial exposed trace is not a complete view of the search system’s internal algorithm.
What to do next
Use the trace as observed context and keep unobserved internal steps unknown.
Measurement scope and aggregation ↗ · Evidence and availability ↗