Accuracy change tracking
Monitor how AI’s description of the business shifts over time and flag unexpected changes.
Glossary
AI termsworth knowing
Plain-language definitions for the terms behind AI in general, AI search, answer engines, and how brands show up in LLM-generated results.
A
Action macros
A set of predefined actions executed with a single one-click operation.
AEO (Answer Engine Optimization)
The practice of optimizing content so AI answer engines surface it accurately.
Agentic AI
AI systems that autonomously plan and execute multi-step tasks toward a goal.
AI agent
A software tool, powered by an AI model, that performs tasks or takes actions on a user’s behalf.
AI alignment
The broader effort to ensure an AI system’s behavior matches human intentions and values, not just technical accuracy.
AI answer checker
A comparison between AI’s answer and your expected answer via a live query.
AI bias
Bias in an AI model’s outputs, often inherited from poor training data, producing unfair or inaccurate results.
AI citations
The references and sources AI systems cite when using your content in an answer.
AI content detection
Tools and techniques used to determine whether a piece of text was generated by an AI model or written by a human.
AI governance
The policies, oversight, and accountability structures a company puts around its use of AI systems.
AI hallucination
AI-generated information that sounds convincing but isn’t actually true.
AI mentions
The frequency and context in which AI systems mention the business.
AI monitoring
See what AI systems are actually saying about your business through citations and trending prompts.
AI Overviews
AI-generated summaries that appear at the top of Google search results, synthesizing information from multiple sources without requiring a click-through.
AI Quality Score
A score measuring structured-data quality across completeness, accuracy, compliance, and performance from crawl and schema validation.
AI Representation Score
A score measuring whether AI engines surface the business in its category, how prominently, and how favorably, through live LLM probing over time.
AI search engine
A search platform that uses AI to generate a synthesized answer from live web results, instead of just returning a ranked list of links.
AI Trust Score
A score measuring how accurately AI engines describe the business, judged against verified claims, evidence, and entity completeness.
AI visibility
How prominently and accurately a brand shows up across AI-generated answers — the AI-era equivalent of showing up on page one of search.
AI-ready data feed
A structured, publicly accessible feed generated from crawled schema and verified claims for AI-platform ingestion.
API access
Programmatic Application Programming Interface (API) access to platform data and actions for external systems.
Authority score
A metric estimating how much an AI model trusts a particular source, influenced by factors like citation frequency, domain reputation, and content quality. *similar to LLMSource’s AI Trust Score
Automated badge revocation
Badge revoked when score, crawl, payment, or security triggers are met.
B
Blocker reports
A list of items blocking certification submission, shown on a failure screen with steps to fix them.
Break-glass override
An emergency access path used when the normal allowed-IP path is unavailable.
BYOK
Bring Your Own Key (BYOK): an option to bring your own encryption key instead of using a platform-managed one.
C
Certification tiers (Bronze/Silver/Gold)
Certification levels according to a domain’s overall score.
Certification workflow
A pre-certification checklist for the issuance of a signed, publicly verifiable badge.
Chain-of-thought reasoning
When a model works through a problem in visible intermediate steps before giving a final answer, rather than jumping straight to a conclusion.
ChatGPT
OpenAI’s ChatGPT, one of the monitored LLM sources for mentions and citations.
Citation frequency
How often AI models cite a given source or domain when answering — higher frequency signals the model treats that source as trustworthy and quotable. *similar to LLMSource’s AI Trust Score
Claude
Anthropic’s Claude, one of the monitored LLM sources for mentions and citations.
Closed-source model
A model whose underlying weights and architecture are kept private by the company that built it, with access only through their own product or API.
Co-citation
Your brand is mentioned near another brand or topic, online or in an AI answer, without an actual link. It still counts as a visibility signal, just a weaker one than a direct citation.
Competitor comparison
Benchmark how AI represents your brand against up to ten competitor domains with gap analysis.
Completeness check
Score how complete structured data is, per page and per schema type.
Compliance dashboard
A checklist of governance requirements with status indicators.
Consistency check
Check whether facts stated in JSON-LD also appear consistently in crawlable body text.
Consistency scoring
Identify facts that AI answers correctly only some of the time across repeated queries.
Content provenance / watermarking (C2PA)
Coalition for Content Provenance and Authenticity (C2PA): a technical standard for embedding verifiable metadata showing whether content was AI-generated, edited, or captured by a camera.
Content scan
Scan page body text beyond schema to check what AI systems would actually read or cite.
Context window
The maximum amount of text (measured in tokens) a model can process in a single conversation or request before it starts “forgetting” earlier parts.
Contradiction flagging
Flag contradictions between schema claims and body content and feed them into the issues list.
Conversational search
Searching via natural, multi-turn dialogue with an AI assistant, rather than typing isolated keywords into a search bar.
CPC (Cost-Per-Click)
The average cost paid per click in correlated advertising performance data.
Crawlability
Indicate whether AI crawlers can access and read a page’s content.
D
Data residency
The storage region assigned to a tenant’s data (EU, US, or APAC), configurable at the child-organization level.
Dedicated infrastructure
Isolated, tenant-specific hosting resources provisioned instead of a shared infrastructure.
Deepfake
Synthetic image, audio, or video content generated by AI to convincingly depict a real person saying or doing something they didn’t.
DeepSeek
One of the monitored LLM sources for mentions and citations.
Domain authority
A third-party metric estimating a domain’s likelihood of ranking well, based on link profile and other signals.
Drift detection
Alert triggered if AI gets previously accurate facts wrong, based on a comparison of scores across scheduled scans.
E
E-E-A-T
Google’s Experience, Expertise, Authoritativeness, and Trustworthiness framework for evaluating content quality. It’s also referenced in discussions of what makes content citable by AI systems.
Entity mapping
Map parent-child relationships between organizational entities.
Explainability (XAI)
The degree to which an eXplainable AI system’s decision-making process can be understood and explained by humans, rather than functioning as an opaque “black box”.
F
Fact accuracy
Comparison of AI-generated descriptions of the business against verified claims with divergences highlighted.
Feed generation
Assemble the AI-ready data feed from crawled schema and verified claims.
Feed publishing
Publish the feed to a public URL and track submission status per platform.
Few-shot learning
A model performing a task with just a couple of examples rather than needing extensive training data for that specific task.
Fine-tuning
Further training an already trained model on a narrower dataset to specialize its behavior for a particular task, tone, or domain.
Foundation model
A large-scale AI model (e.g., GPT, Claude, Gemini’s underlying models) trained on broad data and designed to be adapted for many different downstream tasks or products.
G
Gemini
Google’s Gemini, one of the monitored LLM sources for mentions and citations.
Generative AI
AI systems that create new content — text, images, code, or audio — by learning patterns from training data and producing novel outputs that follow those patterns.
GEO (Generative Engine Optimization)
The practice of optimizing content and brand presence for generative AI systems’ outputs.
Geographic reporting
A performance report broken down by geographic areas.
Google SGE
Google’s Search Generative Experience — an earlier interface using generative AI to provide conversational, synthesized answers within search results, since rebranded as Google AI Overviews.
GPT (Generative Pre-trained Transformer)
A family of large language models developed by OpenAI, pre-trained on vast text corpora to generate human-like text.
Grok
X’s Grok, one of the monitored LLM sources for mentions and citations.
Grounding
The process of connecting an AI model’s outputs to verified, real-world data sources, so its responses are factual rather than speculative or hallucinated.
Guardrails
Rules and safeguards built into an AI system to prevent harmful, inaccurate, or off-brand outputs.
H
Human-in-the-loop
A workflow design where a human reviews, corrects, or approves an AI system’s output before it’s finalized or acted on, rather than letting the AI act fully autonomously.
I
Immutable audit log
An append-only, tamper-proof event log across ten categories that can never be edited or deleted.
Incident runbooks
Step-by-step response workflows for crawl failures, score drops, integration errors, and security incidents.
Inference
The process where an AI model turns your input into tokens, runs them through its trained layers, and predicts the response one token at a time until it reaches an answer.
IP allowlisting
Restrict platform access to an approved list of Internet Protocol (IP) addresses, with emergency override available.
J
JSON-LD
JavaScript Object Notation for Linked Data (JSON-LD): the common code format used to write structured data, placed in a page’s script tag so AI and search engines can read it directly.
K
Knowledge graph SEO
Structuring and verifying entity information so it’s correctly represented in Google’s Knowledge Graph and other structured knowledge graphs, which AI systems often draw on for factual answers.
L
Latency
How long a model takes to respond to a request.
LLM (Large Language Model)
An AI model trained on vast amounts of text to predict and generate language.
LLMO (Large Language Model Optimization)
Making a brand’s content easy for LLMs to parse, retrieve, and represent faithfully — lowering the odds of it being ignored or described incorrectly.
Long-tail query
A longer, more specific search phrase as opposed to a short, broad keyword.
M
Machine learning
The general approach of training a system to improve at a task from data, rather than being explicitly programmed for it.
Metadata
Descriptive data attached to pages or schema that informs AI interpretation.
Microsoft Copilot
Microsoft’s Copilot, one of the monitored LLM sources for mentions and citations.
Missing citations
A highlight of queries where AI should cite the business’s content but doesn’t.
Model weights
The learned numerical values inside a model, locked in after training, that determine how it responds to any input.
Multimodal AI
AI systems capable of processing and generating multiple types of content: text, images, audio, and video.
N
Neural network
The layered mathematical structure that LLMs and most modern AI models are built from, loosely modeled on how brain neurons connect.
NLP (Natural Language Processing)
The broader field of AI focused on understanding and generating human language.
O
Off-site signals
Ranking and trust factors that come from sources other than your own site: backlinks, third-party mentions, reviews, and citations that AI and search systems use to judge credibility.
On-site signals
Ranking and trust factors controlled directly on your own website: schema, page content, and site structure.
Open-source model
A model whose underlying weights and architecture are made publicly available, so anyone can inspect, modify, or build on it.
P
Parameter count
The number of adjustable values inside a model that were tuned during training — often used as a proxy for how large or capable a model is.
Perplexity
Perplexity AI, one of the monitored LLM sources for mentions and citations.
PII masking
Automatic detection and masking of Personally Identifiable Information (PII) in schema and crawl outputs.
Platform layers (Discover/Trust/Control)
Three product layers, each tied to a score: Discover (AI Quality Score), Trust (AI Trust Score), and Control (AI Representation Score).
Policy-as-code
Governance rules written as code that trigger automated actions, such as blocking deployment below a quality threshold.
Product monitoring
Tracking of product schema health, product catalog management, and quality-score impact monitoring.
Prompt engineering
The practice of carefully wording instructions given to an AI model to get more accurate, relevant, or well-formatted responses.
Prompt injection
A security risk where hidden instructions embedded in content trick an AI system into ignoring its original task.
Public badge verification
A public, no-login page confirming a domain’s certification and scores are genuine via a cryptographic signature.
R
RAG (Retrieval-Augmented Generation)
A technique where a model retrieves relevant external content at the moment of answering, rather than relying solely on what it learned during training.
Red teaming
Deliberately stress-testing an AI system for weaknesses, biases, or safety failures before it ships.
ROAS automated correlation analysis
Automatic correlation of AI-visibility changes with Return on Ad Spend (ROAS) performance trends.
ROAS correlation (raw data)
The raw Return on Ad Spend (ROAS) data used before correlation analysis is applied.
Robots.txt review
A check of the site’s posted rules for what crawlers can access, catching rules that accidentally block AI crawlers that hurt the business’s visibility.
S
Scheduled test queries
A saved library of test queries that run on schedule and notify admins when results diverge from expectations.
Schema
A standardized way of organizing information on a site so AI can understand and surface the brand accurately.
Schema detection
Identify existing structured data markup present on a crawled page.
Schema editor
A tool to create or modify structured data markup directly within the platform.
Schema validity
Check whether structured data conforms to required syntax and format standards.
SCIM 2.0 provisioning
System for Cross-domain Identity Management (SCIM) 2.0 provisioning lets you automatically create, update, or remove user accounts to match your company’s identity system, so access always stays in sync.
Semantic search
A retrieval method where a system matches your query to relevant content based on meaning, not exact wording.
Semantic SEO
Semantic SEO (Search Engine Optimization) is the content strategy of writing around a topic comprehensively, not just one target keyword, so both search engines and AI systems recognize your page as a strong match for the broader subject.
Sentiment analysis
Judges whether AI-generated mentions of a brand are positive, neutral, or negative, so you can catch and correct narrative drift early.
SERP
Search Engine Results Page (SERP): the page displaying results after a search query, now often including AI Overviews alongside traditional links.
Share of voice
Your brand’s slice of all AI mentions in a topic or category, measured against competitors — a read on who the models surface first.
Site scan
Crawl the website to find all publicly accessible pages for schema scanning and scoring.
SLO dashboards
Service Level Objectives (SLO): dashboards that track whether operational targets (like how fast an issue gets fixed) are being met, and flag it when they’re breached.
SOC 2 audit export
Export of Service Organization Control 2 (SOC 2) audit data as a structured PDF for external auditors to review in one click.
SSO (SAML 2.0)
Single Sign-On (SSO) using Security Assertion Markup Language (SAML) 2.0 is a protocol that lets a user log into multiple applications with one set of credentials, your company account — no separate password needed.
Structured data
A standard set of tags added to a page that label what each part means, like a business name or price, so AI can read and reuse the content accurately.
Synthetic data
Data generated by an algorithm rather than collected from real-world events, often used to train or test AI models when real data is scarce or sensitive.
T
Tenant encryption
Encryption applied separately to each tenant’s data to maintain isolation and security.
Token pack
A one-time top-up of AI tokens upon exceeding LLMSource’s plan allowance (no rollover of tokens).
Tokens
A unit of text (roughly a word or part of a word) that AI models use to process and generate language. AI usage is often measured in tokens.
Top-K
A parameter controlling how many of the highest-ranked retrieved results (e.g., top 5, top 10) a model considers before generating an answer.
Top-P
A setting that controls how varied an AI’s word choices are — lower values keep answers predictable, higher values allow more creativity.
Trust signals
The cues AI uses to judge reliability: verified claims, consistent page content, mapped entities, and a current data feed.
Two-person rule
Control requiring a second user to approve any changes for accountability.
Type identification
Determine which schema type (e.g., Organization, Product, FAQ) applies to a given page.
U
Unified score dashboard
A single consolidated view of all scores (AI Quality Score, AI Trust Score, and AI Representation Score) for Business and Enterprise plans.
V
Vector embeddings
Numeric representations of words or content that capture meaning, allowing systems to match queries to relevant content based on similarity rather than exact wording.
W
Webhook notifications
An automatic message sent to another system the moment something happens on the platform, like a scan finishing or a score dropping, so you don’t have to keep checking.
Webhook signatures (HMAC)
Hash-based Message Authentication Code (HMAC) signatures attached to every webhook message, so the system receiving it can confirm the message really came from the platform and wasn’t faked or altered.
White-label
The ability for Agencies to rebrand the platform under their own name and visual identity.
Z
Zero-click search
A query answered right in the results or by an AI assistant, with no click-through to any website.
Zero-shot learning
A model performing a task with no examples given (zero-shot) instead of needing extensive training data for that specific task.