Glossary
Clear definitions for the metrics, methods, and terminology behind AI visibility and AI advertising.
A
Active query simulation
Sending constructed queries to an AI service and recording the resulting answers.
Ad creative
The advertiser’s message and assets, including headline, description, image, and destination.
Ad eligibility
Whether an ad satisfies the applicable conditions to compete for delivery. Approval and activation do not guarantee impressions.
Ad group
A collection of ads sharing bidding configuration and context hints.
Ad template
A creative structure populated with details from eligible catalog products.
Ads personalization
Use of eligible signals beyond the current conversation when the user’s settings permit it.
Advanced matching
Eligible first-party information used to improve event matching when direct click information is incomplete.
Agentic commerce
Shopping tasks performed by an AI agent on a person’s behalf. Tasks range from discovery and comparison to checkout assistance and authorized transactions.
AI crawler
Automated software fetching web content for a defined AI-related purpose. Search, training, and ad review can use different crawlers.
AI Mode
Google’s conversational AI search experience. It is a separate experience from AI Overviews, and its answers and sources can differ.
AI Overview
Google’s named AI summary experience in Search. It is not a generic label for every chatbot answer.
AI referral traffic
Visits arriving from identifiable AI services. It measures observed visits, not all exposure or influence inside those services.
AI visibility
AI visibility is how often and how prominently a brand, product, or source appears in AI-generated answers within a defined set of questions and conditions. It includes presence, representation, and citations—not just website visits.
AI-powered text customization
An optional feature adapting existing ad text to conversational context and the user’s language.
Aided awareness
How a model discusses a brand after its name is supplied.
Answer convergence in IQRush measurement
Assessing whether measured answer patterns and estimates become sufficiently consistent as evidence accumulates.
It helps determine whether the available evidence supports a conclusion or more measurement is needed.
Answer engine
A service that responds to questions with generated answers, sometimes using search and citations. Measure the specific service and experience, not only the underlying model name.
Answer engine optimization or AEO
A related industry label for making information easier for answer-producing systems to find and use. Vendors don’t all use AEO and GEO the same way, so check how each defines its scope.
Attention proxies
Indirect signals such as placement or citation interaction used to approximate attention. They are not validated substitutes for actual attention or business outcomes.
Attribution
Rules for assigning credit to an ad interaction for a subsequent outcome.
Attribution Clarity
How explicitly an answer identifies the publisher whose material it uses. It concerns source credit, not advertising conversion attribution.
Attribution window
The eligible time interval between an ad interaction and an outcome. Windows differ for clicks and views and depend on the platform setting.
Audience bid multiplier
A bid adjustment for matching users; unlike audience inclusion, it does not by itself exclude everyone else.
Audience exclusion
A rule making matched users ineligible. Exclusion takes precedence over inclusion.
Automatic advanced matching or AAM
Pixel-based detection, normalization, and hashing of supported customer information for matching. It requires applicable consent handling.
Average position
Mean mention or recommendation position among responses where the brand appears. State how missing brands and unordered answers are handled.
B
Balanced analysis in IQRush measurement
Evaluating responses across a defined range of relevant questions, intents, and audience conditions.
It reduces reliance on a narrow set of favorable questions and makes the scope of conclusions clearer.
Baseline
The documented starting measurement used for later comparison. Preserve comparable conditions or disclose changes.
Baseline reset or re-baselining
Establishing a new reference series after a change breaks historical comparability; retain and label the earlier series.
Bid Cap for conversion optimization
A conversion-oriented bid control used with predicted conversion likelihood. It is not a guaranteed achieved acquisition cost.
Billing event
The chargeable event, such as an impression or valid click. Optimizing for conversions does not mean paying per conversion.
Brand and entity detection in IQRush measurement
Identifying references to the intended brand, products, and aliases in AI responses while distinguishing similarly named entities.
Reliable detection provides the foundation for visibility measurement. Misidentified brands can distort comparisons and priorities.
Brand mention
An identifiable reference to a brand in an answer. It need not include a recommendation or link.
Branded prompt
A prompt that names the target brand or product. It measures responses with that name supplied.
C
Campaign
The object governing objective, budget, schedule, and shared targeting.
Campaign objective
The outcome delivery is intended to optimize toward. It must be distinguished from the event used for billing.
Change detection in IQRush measurement
Identifying differences between comparable measurements and assessing them in light of uncertainty and measurement conditions.
It helps surface changes worth investigating while reducing overreaction to isolated movements. A detected change does not establish its cause.
ChatGPT Ads
OpenAI’s advertising offering in ChatGPT, with separately presented ads and advertiser reporting.
Citation
A reference to a source within an answer. A citation can appear without a mention of the brand that owns the source.
Citation coverage
Percentage of eligible responses citing a specified domain or page at least once.
Citation Decay Rate
Change in citation frequency over time for content or a publisher portfolio. Interpret it with collection intervals and model or index changes.
Citation opportunity
A relevant source or topic where useful brand information is absent or incomplete. An opportunity is a research finding, not proof that editing it will change answers.
Citation Rate
Responses explicitly citing the brand or publisher divided by all responses in the defined query set. Includes linked and named source references; excludes merely inferred sourcing. Citation coverage is equivalent only with matching rules.
Citation share
Percentage of counted citations assigned to a specified domain or page.
Click
A recorded interaction with an ad. A click does not establish that a page loaded or a purchase occurred.
Click-through conversion
A conversion credited to an eligible ad click.
Click-through rate or CTR
Clicks divided by impressions, expressed as a percentage.
Co-mention
Brands appearing together in answers. This is an analytical view of competitive context rather than a standalone proof of substitution.
Comparison query
A question evaluating alternatives against one another.
Competitive Displacement Rate
An emerging measure of answers mentioning one brand but not a specified competitor, divided by all sampled answers. Absence does not prove competitive displacement.
Composite visibility score
A combined index of several metrics. Disclose component values, weighting, normalization, and limitations; the index does not replace its underlying measures.
Confidence interval
A confidence interval expresses uncertainty around a sample estimate using a specified statistical procedure. Under its assumptions, a 95% procedure contains the true parameter in approximately 95% of repeated samples.
Visibility percentages need a measure of precision. An apparent lead can be too uncertain to support a decision. Research: IQRush research on uncertainty in AI visibility.
Confidence intervals in IQRush measurement
Statistical ranges expressing uncertainty around estimated measures under the chosen procedure and its assumptions.
They show the precision of an estimate and help prevent false certainty. A confidence interval does not establish that the underlying AI answer is factually correct.
Consumer share
A stated measure of platform use in the target market that can inform inclusion and aggregation weights.
Content chunk
A self-contained passage that can be retrieved or processed separately. Preserve the context needed to interpret it accurately.
Content clarity
Writing that states the subject, claim, and supporting context plainly.
Content Utilization Rate
Content utilization rate assesses how substantially an AI answer uses a publisher’s material, beyond simply listing a citation. Evaluation compares the answer with the source using a documented method.
Content use and citation credit are different outcomes. Identifying a citation does not establish deep use, and similar wording alone does not prove that a system retrieved a particular page.
Context hint
Additional information about an offering that helps OpenAI assess when an ad is relevant. It is supplied at the ad-group level. A good hint describes a genuine customer need, product attribute, or use situation. A hint is neither an exact-match keyword nor a command controlling delivery. Source: OpenAI context hints.
Context hint list
The set of hints attached to one ad group. OpenAI documents a maximum of 2,000, not a recommended target.
Context window
The amount of information a model can process within an interaction, measured in tokens. It is not a guarantee that every included detail will be used correctly.
Contextual relevance
The fit between an offering and the current conversational need, assessed using multiple signals.
Convergence
Estimates becoming sufficiently stable as observations accumulate under defined criteria. Stable results can still be biased.
Convergence sufficiency in IQRush measurement
Whether the observed convergence is adequate for the particular decision and measurement scope.
The evidence needed for an exploratory content idea may differ from the evidence needed to redirect a substantial budget.
Conversational ad
General editorial label for advertising that supports interaction through conversation; not every ChatGPT ad is a Sponsored Agent.
Conversion
An advertiser-defined action eligible for campaign reporting, such as a completed purchase or registration.
Conversion event setting
Configuration selecting which received event counts for campaign measurement. Creating a setting does not send an event.
Conversion optimization
Delivery optimized toward one selected supported standard event. Confirm billing options in the applicable account and interface.
Conversions API or CAPI
A server-to-server method for sending conversion events to OpenAI.
Cost per action or CPA
Spend divided by attributed conversions, using a stated attribution basis.
Cost per click or CPC
Spend divided by clicks.
Cost per thousand impressions or CPM
Spend divided by impressions, multiplied by 1,000.
Custom audience
A group matched from advertiser-provided customer or prospect identifiers. Uploaded records and matched users are different counts.
Custom event
An advertiser-named action outside the standard taxonomy. Current conversion-optimization guidance excludes custom events as optimization goals.
D
Daily budget
A campaign spending control for a day, subject to platform rules. It does not promise that the budget will be spent.
Data collection architecture
The basis of evidence collection: active simulation, passive observation, platform-native reporting, or a disclosed hybrid.
Data collection methodology
How requests are issued and answers captured, including cadence and retrieval configuration.
Data source
The configured source receiving browser or server events.
Data validation
Checks on accuracy, comparability, sample integrity, and bias before reporting.
Decision threshold
A predeclared rule for taking action based on effect size, uncertainty, cost, and risk.
Decision thresholds in IQRush measurement
Rules for determining when evidence is sufficient to support a specified action, considering uncertainty, effect size, cost, and risk.
They make the connection between measurement and action explicit. They are not hidden confidence thresholds inside an AI platform.
Decision-grade measurement
Decision-grade measurement is evidence with the coverage, validation, reproducibility, and precision needed for a consequential business decision. The appropriate standard depends on the action: changing a substantial budget requires stronger evidence than identifying a topic to investigate.
Decision-grade measurement in IQRush measurement
Decision-grade measurement is evidence with the coverage, validation, reproducibility, and precision needed for a consequential business decision. The appropriate standard depends on the action: changing a substantial budget requires stronger evidence than identifying a topic to investigate.
It helps teams decide what to change, what to test, and when to gather more evidence before committing resources.
Directional measurement
Evidence suitable for trends, early signals, and competitive awareness, with limitations that prevent treating it as sufficient on its own for high-stakes allocation.
Drift detection
Monitoring for shifts beyond normal variation and investigating possible platform, market, or measurement-design causes.
E
Enhanced disclosure
More complete methodological and validation information needed to assess consequential uses of measurement.
Entity clarity
Identifying brands, products, and organizations unambiguously and consistently.
Entity resolution
Matching brand names, products, and aliases to the intended entity while excluding lookalikes. Misidentification can contaminate both numerator and denominator.
Event deduplication
Recognizing duplicate reports of one action so it is not counted twice. Coordinate browser and server event names and IDs.
Event ID
An identifier for an individual event, reused for retries or the same event sent through multiple integrations.
Event receipt
Confirmation that an event reached the measurement source. Receipt is not proof it was attributed to a campaign.
Exploratory measurement
Exploratory measurement is an initial investigation used to generate questions and identify possible patterns. It precedes a sufficiently developed directional or decision-grade assessment.
F
Factual accuracy
Whether assessed claims agree with reliable reference evidence. Report unverified claims separately rather than silently treating them as true.
Factual accuracy classification
Separating accurate, fabricated, source-inaccurate, and unresolved claims against identified reference evidence.
Factual Inaccuracy Rate
For brands, mentions carrying materially wrong source-derived facts divided by total mentions. For publishers, valid citations misrepresenting the cited work divided by total citations. These track-specific meanings are separate from hallucination and are reported by platform.
Fixed bid
A strategy using the advertiser’s specified bid control.
Framing
The role or category assigned to a brand, such as specialist, economical option, or market leader. It is distinct from positive or negative sentiment.
G
Generative answer
A response assembled by a generative AI system. It may combine retrieved material with model knowledge.
Generative engine optimization or GEO
Work intended to improve how accurately and usefully a brand or its content is represented in generated answers. Results require measurement; publication alone does not establish improvement.
Geographic and temporal controls
Documented treatment of location, language, collection timing, and measurement windows to preserve comparability.
GPTBot
OpenAI’s crawler for material that may be used in foundation-model training. Its controls are separate from search controls.
Grounding
Connecting an answer to supporting information. A source should substantiate the associated claim; the presence of a link alone does not prove it does.
H
Hallucination
Generated information that is false or unsupported but presented as factual.
Hallucination Rate
For brands, fabricated or unsupported brand mentions divided by total mentions. For publishers, fabricated attributions divided by total citations. Report the relevant denominator and platform; retain flagged records for review.
Historical data and versioning
Preserving measurement history with records of relevant model, query, and baseline changes.
Holdout or control group
A comparison group not receiving the tested intervention, used to help estimate its effect.
Hybrid measurement architecture
A disclosed combination of collection approaches; their outputs are not assumed interchangeable.
I
Implied citation
Suspected source use without explicit attribution. Excluded from the IAB Citation Rate definition.
Impression
A recorded ad display under the platform’s counting rules.
Incrementality
The outcome caused by an intervention beyond what would otherwise have happened. Attribution alone does not establish it.
Informational query
A question seeking explanation or knowledge.
Intent
The task or decision a person is trying to complete, such as learning, comparing, troubleshooting, or buying.
K
Knowledge cutoff
The boundary of a model’s training knowledge, distinct from information obtained through tools or retrieval.
L
Landing page
The destination reached from an ad; its content also helps inform ad relevance.
Lift
Change relative to a baseline or comparison group. Reports should state whether lift is an absolute percentage-point change or a relative percentage change.
Linked citation
An explicit source reference with a hyperlink.
llms.txt
A proposed Markdown convention helping agents navigate a site’s useful information. It is not evidence of a ranking benefit or guaranteed citation.
M
Margin of error
A stated measure of interval width around an estimate. It does not account for every source of bias or measurement error.
Market-driven shift
A change associated with brand, competitor, or market activity. A measurement pattern alone does not establish the cause.
Maximize Results
Automated bidding represented by supported maximize-clicks or maximize-conversions strategies; availability depends on configuration.
Maximum bid
An upper bid control whose unit depends on campaign and bidding configuration. It is not necessarily the average price paid.
Measurement bias
Systematic distortion caused by selection, setup, extraction, or scoring. Additional observations alone do not remove it.
Measurement conditions
The platform, model or experience, region, language, date, session context, and settings defining a test.
Measurement consent
Permission handling that determines whether measurement events may be sent. The Pixel provides controls; it does not decide an advertiser’s legal obligations.
Measurement explainability in IQRush measurement
Connecting a reported finding to the observed answers, entities, sources, and measurement context that support it.
Teams can inspect the basis for a recommendation rather than relying on an unexplained score.
Measurement stability in IQRush measurement
Consistency of measured results across comparable repeated samples or time windows.
Stability helps teams judge whether a reported result is dependable enough to inform action.
Mention coverage
The percentage of eligible sampled responses containing the brand at least once. State which responses qualify for the denominator.
Mention detection attribution and sentiment logic
Rules for recognizing entities and source credit and classifying sentiment or framing, including ambiguous cases.
Mention Rate
Responses mentioning a brand divided by all responses in the defined query set. Equivalent to mention coverage only when the same denominator and counting rules apply.
Mention share or share of voice
The brand’s share of counted mentions across the comparison set. Reports should state whether repeated mentions within one answer are counted.
Methodology documentation
An account of query selection, collection, scoring, platform handling, aggregation, and limitations sufficient to interpret results.
Minimum disclosure
Summary information needed to place a provider’s measurement in context for directional use.
Model Context Protocol or MCP
An open standard connecting AI applications to external tools and data. It provides connectivity, not an advertising placement or a guarantee of correct results.
Model drift
Changes in AI response behavior over time that can alter measurements independently of brand activity.
Model update impact assessment
Investigation of release information and changes across platforms, brands, and intent types following a suspected update.
Modeled conversion
An attribution estimate using aggregate observed patterns where that measurement is available. Reported totals can include modeling.
Multi-platform aggregation
Combining results across AI services using disclosed weights while retaining separate platform results.
Multimodal visibility
Brand or source representation across text, images, audio, video, and mixed response formats. Measurement standardization is still developing.
N
Named citation
An explicit source reference identifying the source without a hyperlink.
New platform entry protocol
Predetermined rules for adding an AI service while documenting coverage changes and preserving comparable historical reporting.
Noise floor
The noise floor is the ordinary variation in a measurement process against which an apparent change is assessed. It helps explain why a small movement can appear in a dashboard without indicating a meaningful change.
Non-determinism
The possibility of different outputs when a query is repeated under apparently comparable conditions.
Nonbranded prompt
A prompt that does not supply the target brand or its identifying aliases. Useful for studying unaided discovery.
O
OAI-AdsBot
OpenAI’s crawler for reviewing submitted ad landing pages and assessing relevance; its collected data is not used for foundation-model training.
OAI-SearchBot
OpenAI’s crawler for surfacing websites in ChatGPT search.
Observed versus estimated fan-out
Observed fan-outs come from a captured search trace; estimated fan-outs are hypotheses about likely searches. Label the two separately.
oCPC
Conversion-oriented optimization with valid-click billing. The optimization goal and payment unit differ.
OpenAI Measurement Pixel
Browser software that sends website measurement events for ChatGPT Ads.
oppref
OpenAI’s opaque click reference, added to the landing URL and used for attribution matching. Preserve its original value.
Organic AI visibility
Unpaid presence in an AI answer or its supporting references. It is distinct from separately labeled advertising.
Owned source
A source controlled by the brand, such as its website.
P
Paid AI advertising
Paid placements in AI experiences. Ads are labeled separately from independently generated answers.
Panel validity
How well an observed participant sample supports conclusions about the target population, considering recruitment, size, incentives, and representation.
Passive behavioral observation
Observing consenting participants’ actual interactions through a panel.
Persona
A research description of a type of buyer and their needs. A simulated persona is not proof of a real user’s identity or behavior.
Persuasion
The endorsement and action dimension of visibility, assessed through recommendation strength and available click evidence. This does not by itself establish causal impact.
Pixel ID
The identifier used by measurement integrations to associate events with a source; server-only integrations also use it.
Platform coverage
The AI services and versions included in the measurement scope.
Platform-driven bias
Systematic differences in how an AI service represents brands or sources, assessed separately from prompt-selection effects.
Platform-driven shift
A change associated with the AI service, such as retrieval, model, or answer-presentation changes.
Platform-native data
First-party reporting supplied by the AI platform.
Point estimate
A single numerical estimate from a sample, such as 30% mention coverage.
Portrayal
How an answer characterizes the brand or publisher, including context and correctness.
Position
Placement of a brand in the rendered answer, including order, depth, repetition, and standalone versus shared recommendation. Average position is only one possible summary.
Post-Citation Click-Through Rate
The share of brand mentions or publisher citations that lead to the corresponding digital property. Brand and publisher denominators are calculated separately. Requires eligible platform exposure and click data; referral sessions alone are insufficient.
Post-click conversion rate
Click-attributed goal conversions divided by clicks.
Practical significance
Whether a measured difference is large enough to matter for the business decision.
Presence
Whether a brand appears or a publisher is explicitly cited in the measured answers.
Probabilistic output
An output that can vary across runs or conditions. The same question need not yield the same set or ordering of brands.
Product feed
Structured catalog data used to populate product ads.
Product set or product filter
Rules selecting which catalog items an ad group can advertise.
Prominence
How visibly and substantially the brand or source appears within an answer.
Prompt
The input presented to an AI system. Record the exact wording and relevant context when using it as a measurement unit.
Prompt injection
Instructions embedded in untrusted content that attempt to redirect an AI system’s behavior. It differs from legitimate editorial practices that make information clearer.
Prompt library construction
How a maintained collection of measurement queries is assembled and refreshed.
Prompt set
The documented collection of questions used for a study. It should reflect the scope being measured rather than only questions likely to favor the brand.
Prompt type coverage
Distribution across informational, comparison, recommendation, and transactional intents. Report the mix and results by intent.
Prompt volume
An estimate of user demand for a question or topic. Identify the dataset, period, geography, and estimation method; test-run counts are not consumer demand.
Prompt-driven bias
Systematic distortion caused by an unbalanced selection of question types, wording, or intents.
Q
Query fan-out
Related searches generated to gather information for an answer. These searches can differ from the user’s original wording.
Query set transparency
Disclosure of query construction, selection, grouping, weighting, intent mix, and refresh practices.
Query sourcing
Whether queries originate from constructed research prompts, search-demand data, observed behavior, or a combination.
Query volume
Number of queries issued in a measurement program. Report it separately from responses per query and from estimated consumer prompt demand.
R
Rank stability
Rank stability is the consistency of relative ordering as comparable evidence is collected. IQRush’s research evaluates stability alongside structural sufficiency before interpreting differences in citation visibility.
A brand moving from second to first may reflect sampling variation rather than a durable advantage. Read more: How to tell if your AI search visibility number is real.
Rank stability in IQRush measurement
Rank stability is the consistency of relative ordering as comparable evidence is collected. IQRush’s research evaluates stability alongside structural sufficiency before interpreting differences in citation visibility.
A brand moving from second to first may reflect sampling variation rather than a durable advantage. Read more: How to tell if your AI search visibility number is real. It helps avoid treating a small, unstable ranking difference as a durable competitive advantage.
Recency bias
A tendency to favor newer information. Consider its possible effect on comparisons involving content of different ages.
Recommendation query
A question seeking a suitable option for a stated need.
Recommendation rate
The percentage of eligible responses that recommend the brand, using a documented rule to distinguish endorsement from incidental mention.
Recommendation Strength
The degree of endorsement in an answer, including specificity, prominence, and qualifications. A strong endorsement differs from an incidental mention; disclose the scoring rubric.
Relevance-weighted second-price auction
OpenAI’s stated auction approach incorporates relevance when selecting eligible ads. The charge is not necessarily the next advertiser’s unadjusted bid plus a fixed increment.
Remediation in IQRush measurement
Prioritized changes intended to address observed visibility or representation problems, followed by comparable measurement.
It connects diagnosis to practical work and checks whether outcomes improved. Improvement should not be attributed to the intervention without suitable evidence.
Repeated answer measurement in IQRush measurement
Collecting multiple AI responses under documented conditions to characterize variation.
A single answer is a snapshot. Repeated observations help distinguish recurring patterns from isolated outcomes.
Repeated sampling
Collecting multiple responses under comparable conditions to estimate output variation. Repetition and coverage of different buyer questions serve different purposes.
Representation analysis in IQRush measurement
Assessing how an AI answer describes a brand, including its attributes, recommendations, sentiment, and support for factual claims.
A mention can help or harm a brand depending on what the answer says. Presence alone does not explain the quality of exposure.
Reproducibility
Consistency when comparable queries are rerun, judged against documented variation rather than identical wording.
Retrieval configuration
Whether live external retrieval is enabled and how that condition is recorded for comparison.
Retrieval-augmented generation or RAG
Generating a response with information retrieved from external sources in addition to the model’s learned parameters. Retrieval can improve grounding without guaranteeing correctness.
Return on ad spend or ROAS
Attributed purchase value divided by ad spend. Revenue attribution does not equal incremental profit.
S
Sample size
The number of observations used in an estimate. Unique prompts, repeat runs, completed answers, and independent observations are different counts and should be reported separately.
Sampling uncertainty
Uncertainty caused by observing a sample rather than the complete target population or process.
Semantic similarity
Similarity in meaning rather than exact wording. A similarity score does not establish truth, relevance to every user, or causation.
Sentiment
The assessed positive, neutral, or negative characterization of a brand. Identify the scoring scale and classification method.
Share of Voice
The brand’s portion of counted brand mentions within a declared competitive category. The competitor set and counting rules should be stated up front.
Shopping trigger rate
Percentage of sampled runs that activate a particular shopping experience. It differs from appearing within that experience.
Shopping visibility
Frequency of product presence in a defined sample of AI shopping outputs. A useful measure states the product, retailer, surface, and denominator.
Signal-to-noise ratio or SNR
The magnitude of a defined signal relative to a defined noise measure. Interpret it using the stated signal, noise basis, and measurement scope; there is no universal AI visibility SNR.
Source diversity
The range of distinct domains, pages, or source types represented in a response set. More sources do not automatically mean better evidence.
Source freshness
How current the information in a source is for the question being answered. Recent crawling does not guarantee that the content itself is current.
Source tracking in IQRush measurement
Recording cited sources associated with sampled AI answers and examining patterns across the response set.
It reveals which sources appear in the observed evidence environment and supports investigation of content opportunities. Citation frequency alone does not establish causal influence.
Source type classification
Distinguishing brand-owned, retailer, editorial, and institutional sources.
Sponsored Agent
An ad format that opens a conversation with a business’s AI representative. Currently limited to selected alpha advertisers.
Standard event
An OpenAI-defined action type such as a purchase, completed registration, or paid subscription.
Statistical significance
Evidence against a specified null hypothesis under a statistical test and its assumptions. It does not establish business importance.
Structural sufficiency
Structural sufficiency describes whether differences in citation shares are large enough relative to estimation uncertainty to support comparison. In IQRush’s published research, it complements rank stability: a settled ranking may still contain differences too small to resolve.
It helps prevent a leaderboard from presenting an uncertain comparison as a clear competitive lead. Research: Ronald Sielinski on rank stability and structural sufficiency.
Structural sufficiency in IQRush measurement
Structural sufficiency describes whether differences in citation shares are large enough relative to estimation uncertainty to support comparison. In IQRush’s published research, it complements rank stability: a settled ranking may still contain differences too small to resolve.
A stable ordering is useful only when the evidence also resolves the differences relevant to the decision.
T
Targeting
Explicit eligibility controls, including supported geography, platform, and audience settings. Hints do not replace them.
Testing cadence
How often measurement runs are collected. Match collection frequency to the decision cycle and platform changes.
Thematic clustering
Grouping queries into documented topics or categories.
Third-party source
A source controlled by another party, such as a publisher, retailer, or review site.
Topic
A defined subject grouping related questions. Clear boundaries help prevent duplicate coverage from distorting comparisons.
Topic coverage
Topic coverage is the breadth of relevant questions, intents, and audience needs represented in a measurement design. Broad coverage and repeated observations address different weaknesses: missing questions and uncertain estimates.
Transactional query
A question seeking to complete an action, such as locating a purchase channel.
U
Unaided awareness
Whether a model brings up a brand without being given its name.
UTM parameters
URL labels used by analytics tools to identify campaign traffic. They do not replace OpenAI’s click identifier.
V
Variability baseline
A documented range or distribution of ordinary repeated-measurement variation used to interpret later changes.
View-through conversion
A conversion credited to an eligible ad impression under applicable rules. Availability and aggregation need verification.
Visibility Momentum
Percentage change in Mention Rate or Share of Voice between specified periods. State baseline resets and handle a zero starting value explicitly.
Z
Zero-click discovery
Learning about a brand or product without visiting its website. Website traffic therefore captures only part of discovery.