Xlearn by Sophyx / Glossary

AEO, GEO and AI glossary

All 294 terms our courses teach, each in one plain sentence: how ChatGPT, Gemini and other AI assistants choose which businesses to name, and how to use AI well at work. Every term links to the lesson that teaches it. New to AI search? Start with Answer Engine Specialist. New to AI itself? Start with AI Fundamentals. Both end in a certificate. Your first certification is free.

#

@id
A stable IRI that names a node, so every page can refer to the same thing.
Taught in Structured data as a graph (GEO-300, lesson 4)

A

Access map
One page of every tool and source, at its least access.
Taught in Choose what it reads (AX-300, lesson 6)
Accessibility tree
The names, roles and states a browser exposes; what an agent reads.
Taught in Pages agents can use (GEO-300, lesson 7)
Accessible name
The label an agent uses to find a control, from a label element, its text or aria-label.
Taught in Pages agents can use (GEO-300, lesson 7)
Action item
A task from a meeting, with an owner and a date.
Taught in Research and summaries (AX-100, lesson 10)
AEO (answer engine optimization)
Making your content the clear, direct answer to a specific question.
Taught in From ten blue links to one answer (AE-200, lesson 1)
Agent loop
Plan, act, check the result, then decide the next step.
Taught in What makes an agent (AX-300, lesson 1)
Agent order
An order placed through an agent’s checkout connection.
Taught in Count agent orders (COM-330, lesson 10)
Agent test
Real buyer requests run in AI assistants, picks recorded.
Taught in The agent-ready loop (COM-330, lesson 12)
Agent walkthrough
Watching an agent complete one task on your site and logging where it fails.
Taught in Pages agents can use (GEO-300, lesson 7)
Agent-ready checklist
Your store’s list of what agents need, kept current.
Taught in The agent-ready loop (COM-330, lesson 12)
Agentic checkout
A purchase an agent completes through the shop’s system.
Taught in How agent checkout works (COM-330, lesson 7)
AI agent
AI that works toward a goal and chooses its next step.
Taught in What makes an agent (AX-300, lesson 1)
AI referral
A visit or order from a person who clicked an AI answer.
Taught in Count agent orders (COM-330, lesson 10)
AI referral traffic
Visits that arrive from an AI assistant, seen as referrals in your analytics.
Taught in Measure what wins (AE-200, lesson 11)
AI referral visit
A visit that comes from a link in an AI answer.
Taught in The AI visibility chain (GEO-310, lesson 1)
AI visibility chain
Crawler reads, mentions and citations, visits, conversions.
Taught in The AI visibility chain (GEO-310, lesson 1)
Allowed values
The only answers a field may take, such as yes or no.
Taught in Templates and structured outputs (AX-200, lesson 5)
Alt text
A short text description of an image.
Taught in Beyond text (AEO-210, lesson 9)
Alternatives page
A page listing real options and who each one is best for.
Taught in Honest comparisons (AEO-210, lesson 6)
Analysis tool
A feature that runs real calculations on your file.
Taught in Numbers and spreadsheets (AX-100, lesson 11)
Annotation
A dated note on a chart: what changed, by you or by a platform.
Taught in Dashboards that drive action (GEO-310, lesson 10)
Answer page
A page built around one question, with the direct answer first.
Taught in Page types that win answers (AEO-210, lesson 3)
Answer-first writing
Opening a section with the direct answer, then the explanation.
Taught in Answer-shaped content (AE-200, lesson 6)
API
A door one program offers so other programs can use it.
Taught in Give it the right tools (AX-300, lesson 4)
App shell
An HTML page that is empty until JavaScript fills it; empty to most fetchers.
Taught in What fetchers see (GEO-300, lesson 3)
Approval fatigue
Clicking yes without reading, after too many checks.
Taught in A person in the loop (AX-300, lesson 9)
Approval gate
A pause before an action until a person says yes.
Taught in A person in the loop (AX-300, lesson 9)
Approval record
Saved proof of what the buyer approved, and when.
Taught in Returns and disputes (COM-330, lesson 11)
Approval step
A pause where a person approves, edits or declines.
Taught in Hand-offs and approvals (AX-200, lesson 7)
As-of date
The date a number was true.
Taught in Evidence and specifics (AEO-210, lesson 7), Your official facts hub (REP-320, lesson 5)
Assisted conversion
A conversion where an AI visit was one of the steps, but not the last one.
Taught in From prompt to page to visit (GEO-310, lesson 7)
Astroturfing
Fake grassroots support: paid or fake accounts posing as independent fans.
Taught in Communities, with disclosure (REP-320, lesson 9)
Audit log
A lasting record of who did what, and when.
Taught in Watch every run (AX-300, lesson 10)
Audit sheet
One row per prompt, assistant and claim, with its grade and its source.
Taught in The reputation audit (REP-320, lesson 3)
Author page
A page about the author: role, experience and profiles.
Taught in Authors and expertise (AEO-210, lesson 8)

B

Baseline
Your numbers before the change, measured the same way.
Taught in Measure and improve (AX-200, lesson 10)
Bot alert
A rule that pages an owner when verified bots get errors or a new token appears.
Taught in AI bot logs, verified (GEO-300, lesson 10)
Bot management
Rules that decide which automated visitors get in.
Taught in Let trusted agents in (COM-330, lesson 9)
Buyer approval
The buyer’s yes to the exact item and total, before paying.
Taught in How an agent picks (COM-330, lesson 1)
Buyer’s limits
What the buyer allows: item, price, shop and time.
Taught in Limits, consent and the total (COM-330, lesson 8)
Byline
The name and role of the person who wrote a page.
Taught in Authors and expertise (AEO-210, lesson 8)

C

Change log
A dated record of every policy change, so a drop in fetches can be matched to its cause.
Taught in GEO CI and governance (GEO-300, lesson 12)
Changelog
A short record of what changed on a page and when.
Taught in Freshness workflows (AEO-210, lesson 10)
Channel group
Rules that sort sessions into channels such as organic search, referral or AI assistant.
Taught in Tracking AI referrals (GEO-310, lesson 6)
Channel tag
A label on each order saying where it came from.
Taught in Count agent orders (COM-330, lesson 10)
Chargeback
A payment the bank reverses after the buyer disputes it.
Taught in Returns and disputes (COM-330, lesson 11)
Check-and-cost grid
Can you check it? What would a mistake cost?
Taught in When not to use AI (AX-100, lesson 12)
Checkable detail
A name, number, date, quote or link you can verify.
Taught in Why it gets things wrong (AX-100, lesson 3)
Checkout flag
A feed field saying whether agents may buy the item.
Taught in Feeds for AI shopping (COM-330, lesson 4)
Citation
A link or reference an answer gives for a claim.
Taught in Open the source (AX-100, lesson 8)
Citation share by domain
The share of citations that point to each domain, such as your site, a directory or a review site.
Taught in Benchmarking competitors (GEO-310, lesson 8)
Claim ledger
The list of every claim about the brand, with its type, its source and its status.
Taught in Facts and feelings (REP-320, lesson 2)
Clarifying question
A question it asks you before it writes.
Taught in Show it what good looks like (AX-100, lesson 5)
Clean session
A run with no memory, history or personalisation and a fixed location, so runs are comparable.
Taught in Sampling AI answers (GEO-310, lesson 4)
Co-mention
Two brands named in the same answer.
Taught in Benchmarking competitors (GEO-310, lesson 8)
Community answer
A helpful, disclosed reply in a forum or Q&A site.
Taught in Distribution for AI visibility (AEO-210, lesson 11)
Community rules
A forum’s own rules on promotion and disclosure; they come before your plan.
Taught in Communities, with disclosure (REP-320, lesson 9)
Comparison criteria
The shared factors every option is judged on.
Taught in Honest comparisons (AEO-210, lesson 6)
Comparison page
A page that weighs options on clear, shared criteria.
Taught in Page types that win answers (AEO-210, lesson 3)
Conditional request
A fetch with If-None-Match or If-Modified-Since, answered with 304 when nothing changed.
Taught in Fast, reliable responses (GEO-300, lesson 9)
Confidence interval
The range a rate could plausibly sit in, given how many runs you made.
Taught in Sampling AI answers (GEO-310, lesson 4)
Confidential data
Business information not meant for outsiders.
Taught in Keep private data out (AX-100, lesson 9)
Conflicting fact
One fact with two different values in two places.
Taught in When facts disagree (COM-330, lesson 3)
Content decay
A page losing visibility or accuracy over time.
Taught in Freshness workflows (AEO-210, lesson 10)
Content gap
A cluster where AI answers cite others and you have no strong page.
Taught in From prompt set to content map (AEO-210, lesson 2)
Content inventory
A list of pages with owner, purpose, last update and next review.
Taught in Freshness workflows (AEO-210, lesson 10)
Content map
A table linking each prompt cluster to the page that answers it.
Taught in From prompt set to content map (AEO-210, lesson 2)
Content scorecard
Per-page results: reads, citations, AI visits and conversions.
Taught in Measuring content in AI search (AEO-210, lesson 12)
Content signal
A proposed robots.txt line stating search, AI-input and AI-train preferences.
Taught in Bot policy as code (GEO-300, lesson 2)
Context window
Everything the model can see in this chat right now.
Taught in Where answers come from (AX-100, lesson 2)
Continued influence
A corrected claim still shapes what people believe.
Taught in When old crises linger (REP-320, lesson 10)
Control group
Pages you leave unchanged, so you can see what would have happened anyway.
Taught in Experiments that hold up (GEO-310, lesson 9)
Control token
A robots.txt name that sets how a crawl may be used for AI. It fetches nothing.
Taught in Bot policy as code (GEO-300, lesson 2)
Correction ladder
Your page, then profiles and directories, then the publisher, then the thread, then feedback.
Taught in Fix it at the source (REP-320, lesson 4)
Corroboration
Several independent sources stating the same fact.
Taught in Earned media and experts (REP-320, lesson 8)
Cost per result
A month’s cost divided by the results you used.
Taught in Count the cost per run (AX-300, lesson 11)
Cost per run
What one run spends on the model and its tools.
Taught in Count the cost per run (AX-300, lesson 11)
Crawl capacity
How much a crawler is willing to fetch from you; it falls with slow responses and errors.
Taught in Fast, reliable responses (GEO-300, lesson 9)
Crawler read
An AI crawler fetching a page. It shows in server logs.
Taught in The AI visibility chain (GEO-310, lesson 1)
Custom assistant
An assistant you set up once with instructions and files.
Taught in Build a reusable assistant (AX-200, lesson 4)

D

Dark pattern
A design that tricks people into choices they didn’t mean.
Taught in Limits, consent and the total (COM-330, lesson 8)
Decision point
A step where someone chooses what happens next.
Taught in Map the steps (AX-200, lesson 2)
Decision rule
The metric, the threshold and the date, written before the experiment.
Taught in Experiments that hold up (GEO-310, lesson 9)
Decision table
One row per bot job and token: allow, block or limit, and why.
Taught in Bot policy as code (GEO-300, lesson 2)
Decision-first report
A report that opens with the decisions, then shows the numbers that support them.
Taught in Reporting to leadership (GEO-310, lesson 12)
Difference-in-differences
The change in the test pages minus the change in the control pages.
Taught in Experiments that hold up (GEO-310, lesson 9)
Direct answer
One or two sentences that answer the question completely.
Taught in The answer-first page (AEO-210, lesson 4)
Disambiguation
The facts that tell your entity apart from same-name or nearby ones.
Taught in Entities and knowledge panels (REP-320, lesson 6)
Disclosure
A clear note that the author is one of the brands compared.
Taught in Honest comparisons (AEO-210, lesson 6), Communities, with disclosure (REP-320, lesson 9)
Double down
Create more prompts and pages like the ones already winning.
Taught in The answer loop (AE-200, lesson 12), Measuring content in AI search (AEO-210, lesson 12), Winning prompts (GEO-310, lesson 11)

E

Earned media
Coverage and mentions you did not pay for: reviews, articles, lists, posts.
Taught in Many voices (AE-200, lesson 8)
Earned mention
Independent coverage of the brand that was not paid for.
Taught in Earned media and experts (REP-320, lesson 8)
Edge case
An unusual case that’s likely to trip a workflow up.
Taught in Test before you trust (AX-200, lesson 8)
Entity
A clearly identified thing, such as a business, product or person, that AI can recognise across sources.
Taught in SEO, GEO and AEO: what carries over (AE-200, lesson 3), How AI reads your content (AEO-210, lesson 1), Entities and knowledge panels (REP-320, lesson 6)
Entity node
One JSON-LD object for one real-world thing, identified by its @id.
Taught in Structured data as a graph (GEO-300, lesson 4)
Error alert
A message to a named person when a run fails.
Taught in When it breaks (AX-200, lesson 9)
Escalation ladder
Six rungs, watch to legal, each with an owner and a time limit.
Taught in Watch, escalate, respond (REP-320, lesson 12)
Eval
A repeatable test of the agent on real tasks.
Taught in Test it with evals (AX-300, lesson 7)
Example
A sample that shows the style or shape you want.
Taught in Show it what good looks like (AX-100, lesson 5)
Excessive agency
An agent with more tools, access or freedom than it needs.
Taught in Guardrails that hold (AX-300, lesson 8)
Expert route
Going to a qualified person: a vet, doctor or lawyer.
Taught in When not to use AI (AX-100, lesson 12)
Expert source
A named person with checkable credentials who can be quoted.
Taught in Earned media and experts (REP-320, lesson 8)

F

Fact claim
A statement about the brand that can be checked: a date, a number, a name, a place.
Taught in Facts and feelings (REP-320, lesson 2)
Fact consistency
The same core facts on the site, the profiles and the directories.
Taught in Your official facts hub (REP-320, lesson 5)
Fact owner
The person who keeps one fact right everywhere.
Taught in When facts disagree (COM-330, lesson 3)
Fact sheet
One product’s checkable facts, each with where it lives.
Taught in What an agent reads (COM-330, lesson 2)
Facts
The details it must use, pasted in so it can’t guess.
Taught in The four-part prompt (AX-100, lesson 4)
Facts hub
One plain-text page of the brand’s core facts, each with an as-of date.
Taught in Your official facts hub (REP-320, lesson 5)
Fake-review pattern
A burst, similar wording or new accounts: marks of fakes.
Taught in Fakes, errors and rivals (REP-320, lesson 11)
Fallback
How the job still gets done when the workflow stops.
Taught in When it breaks (AX-200, lesson 9)
Feed field
One labelled slot in a feed, such as price or stock.
Taught in Feeds for AI shopping (COM-330, lesson 4)
Feedback channel
An assistant’s report or thumbs-down button: a signal to its makers, not a correction.
Taught in Fix it at the source (REP-320, lesson 4)
Final total
The full price with tax, delivery and fees, before paying.
Taught in Limits, consent and the total (COM-330, lesson 8)
First-hand experience
Details that come from doing the thing yourself.
Taught in Authors and expertise (AEO-210, lesson 8)
Formula
A spreadsheet instruction that does the sum for you.
Taught in Numbers and spreadsheets (AX-100, lesson 11)
Four-part prompt
Task, context, facts and format, in one request.
Taught in The four-part prompt (AX-100, lesson 4)
Fraud check
Checks on an order for stolen cards or fake buyers.
Taught in Let trusted agents in (COM-330, lesson 9)
Fresh chat
A new conversation that starts clean, with your best prompt.
Taught in Improve it in turns (AX-100, lesson 6)
Freshness
How current a page is, shown by updated content and dates.
Taught in Fresh or forgotten (AE-200, lesson 9), Reading crawler logs (GEO-310, lesson 5)
Freshness check
A regular check that key facts match everywhere.
Taught in The agent-ready loop (COM-330, lesson 12)

G

GEO (generative engine optimization)
Helping AI systems understand, trust and mention your brand in the answers they generate.
Taught in From ten blue links to one answer (AE-200, lesson 1)
GEO check
An automated test in CI that fails the deploy when AI access or readability breaks.
Taught in GEO CI and governance (GEO-300, lesson 12)
Grader
A rule, a rubric or a person that scores a run.
Taught in Test it with evals (AX-300, lesson 7)
Grain
The unit a metric is counted at: a run, a prompt, a page or a period.
Taught in The metric dictionary (GEO-310, lesson 3)
Graph
The set of nodes and the links between them, expressed with @graph.
Taught in Structured data as a graph (GEO-300, lesson 4)
GTIN
The global barcode number for one exact product.
Taught in Identifiers and variants (COM-330, lesson 6)
Guardrail
A check the system enforces on what the agent reads or does.
Taught in Guardrails that hold (AX-300, lesson 8)

H

Hallucination
A confident AI statement that no source supports.
Taught in Fresh or forgotten (AE-200, lesson 9), Fakes, errors and rivals (REP-320, lesson 11), Why it gets things wrong (AX-100, lesson 3)
Hand-off
Passing work to a person with everything they need.
Taught in Hand-offs and approvals (AX-200, lesson 7)
Hard cap
A limit the system enforces, which stops the run.
Taught in Count the cost per run (AX-300, lesson 11)
Heading order
One h1, then h2 and h3 in order, so the outline matches the page.
Taught in Semantic HTML for passages (GEO-300, lesson 5)
Health signal
A number you watch each week, like failed tool calls.
Taught in Watch every run (AX-300, lesson 10)
High stakes
When a mistake could hurt someone or cost real money.
Taught in Why it gets things wrong (AX-100, lesson 3)
House style
How your organisation sounds and lays things out.
Taught in Show it what good looks like (AX-100, lesson 5)
HTML-first
Publishing key content as web text before any other format.
Taught in Beyond text (AEO-210, lesson 9)
Hub page
A page that links all answer pages on one topic.
Taught in Page types that win answers (AEO-210, lesson 3)
Human decision
A call about people or safety that a person makes.
Taught in When not to use AI (AX-100, lesson 12)
Human in the loop
A person checks or decides at a set point in a workflow.
Taught in Hand-offs and approvals (AX-200, lesson 7)

I

Improvement record
Each change, why you made it, and what the evals said.
Taught in Improve it safely (AX-300, lesson 12)
Incentivised review
A review given for a reward: disclose it, and never tie it to the rating.
Taught in Honest review programs (REP-320, lesson 7)
IndexNow
A protocol to push changed URLs to participating search engines with a key file; Google does not take part.
Taught in Discovery files and feeds (GEO-300, lesson 6)
Intent
What the buyer is trying to do: discover options, compare them, or decide.
Taught in Prompts, not keywords (AE-200, lesson 10)
Inverted pyramid
The conclusion first, then the supporting detail.
Taught in The answer-first page (AEO-210, lesson 4)
Iteration
Improving an answer in rounds of specific feedback.
Taught in Improve it in turns (AX-100, lesson 6)

J

Job card
The agent’s standing goal, limits, finish line and budget.
Taught in Write the job card (AX-300, lesson 2)
JSON-LD
The most common format for adding schema.org structured data to a page.
Taught in Speak machine (AE-200, lesson 5)

K

Key event
An event you mark as important, such as a booking; GA4's name for a conversion.
Taught in Tracking AI referrals (GEO-310, lesson 6)
Key-facts box
A short block of plain facts: price, duration, place, who it is for.
Taught in The answer-first page (AEO-210, lesson 4)
Knowledge cutoff
The date after which the model learned nothing new.
Taught in Where answers come from (AX-100, lesson 2)
Knowledge file
A reference document the assistant searches for answers.
Taught in Build a reusable assistant (AX-200, lesson 4)
Knowledge panel
The information box a search engine builds about an entity from sources across the web.
Taught in Entities and knowledge panels (REP-320, lesson 6)
Known unknown
A thing you cannot measure, named in the report so nobody assumes it is zero.
Taught in Reporting to leadership (GEO-310, lesson 12)

L

Language model
Software that writes by predicting likely next words.
Taught in What an assistant really does (AX-100, lesson 1)
lastmod
The date a page last changed in meaning; bots learn to ignore a lastmod that is always "today".
Taught in Discovery files and feeds (GEO-300, lesson 6)
Leading indicator
A link early in the chain, such as reads or citations, that moves before visits do.
Taught in Dashboards that drive action (GEO-310, lesson 10)
Leak
A prompt that is cited in answers but whose pages bring no one.
Taught in Winning prompts (GEO-310, lesson 11)
Least access
Giving a workflow only the access its steps need.
Taught in When it breaks (AX-200, lesson 9)
llms.txt
A proposed plain-text file that points AI systems to a site's most important pages. Support varies by assistant.
Taught in Speak machine (AE-200, lesson 5), Discovery files and feeds (GEO-300, lesson 6)
Long-term memory
Notes saved for later runs, read at the start of each.
Taught in Plans and memory (AX-300, lesson 3)
Longest match
In robots.txt, the most specific matching Allow or Disallow path wins.
Taught in Bot policy as code (GEO-300, lesson 2)

M

MCP
An open standard that plugs tools and data into AI apps.
Taught in Connect with MCP (AX-300, lesson 5)
MCP server
A service that exposes tools and resources to AI applications over the Model Context Protocol.
Taught in APIs and MCP for brands (GEO-300, lesson 8), Connect with MCP (AX-300, lesson 5)
Mention vs. citation
A mention names your brand in the answer; a citation links to a source page. You can have one without the other.
Taught in How an AI answer is made (AE-200, lesson 2)
Merchant of record
The seller who takes the payment and owns the order.
Taught in How agent checkout works (COM-330, lesson 7)
Metric dictionary
One written definition per metric: name, formula, source, grain, period, owner and caveats.
Taught in The metric dictionary (GEO-310, lesson 3)
Minutes a week
Times a week multiplied by the minutes each time.
Taught in Spot the repeat work (AX-200, lesson 1)
Mirror check
The monthly audit of your prompt set against the baseline.
Taught in Watch, escalate, respond (REP-320, lesson 12)
Must-have
A condition the buyer sets that every pick must meet.
Taught in How an agent picks (COM-330, lesson 1)

N

NAP consistency
Your name, address and phone number match exactly across every listing.
Taught in Your digital footprint (AE-200, lesson 7)
Net time saved
Old minutes minus checking, fixing and upkeep time.
Taught in Measure and improve (AX-200, lesson 10)
Never list
Jobs that stay with a person, however well they score.
Taught in Score it before you build (AX-200, lesson 3), Write the job card (AX-300, lesson 2)
New-colleague test
Could someone new do the job from your prompt alone?
Taught in The four-part prompt (AX-100, lesson 4)

O

Offer
The price, stock and condition a product is sold at.
Taught in Facts on the page (COM-330, lesson 5)
OpenAPI description
A machine-readable file that describes your endpoints, inputs and outputs.
Taught in APIs and MCP for brands (GEO-300, lesson 8)
Operating guide
One page on how the agent is run, watched and stopped.
Taught in Improve it safely (AX-300, lesson 12)
Original research
Data you collected yourself, such as a survey or a test.
Taught in Evidence and specifics (AEO-210, lesson 7)
Original source
Where a fact was first published, not a copy of it.
Taught in Open the source (AX-100, lesson 8)

P

Page ledger
One row per page: fetches, freshness, citations, AI sessions and key events, from three sources.
Taught in From prompt to page to visit (GEO-310, lesson 7)
Pass bar
The score a workflow must reach before it goes live.
Taught in Test before you trust (AX-200, lesson 8)
Passage
A heading and the text under it: the unit an assistant quotes.
Taught in Semantic HTML for passages (GEO-300, lesson 5)
Passage (chunk)
A short section of a page that a retrieval system scores and may quote on its own.
Taught in How AI reads your content (AEO-210, lesson 1)
Payment token
A one-time stand-in for a card, capped in amount and time.
Taught in How agent checkout works (COM-330, lesson 7)
Personal data
Anything that identifies a living person, or could.
Taught in Keep private data out (AX-100, lesson 9)
Plain text
Words on the page, not hidden in images or files.
Taught in What an agent reads (COM-330, lesson 2)
Plan
The agent’s written checklist of steps for this run.
Taught in Plans and memory (AX-300, lesson 3)
Policy owner
The named person who approves changes to robots.txt, WAF rules and structured data.
Taught in GEO CI and governance (GEO-300, lesson 12)
Product facts
Checkable details: price, stock, delivery, returns, specs.
Taught in What an agent reads (COM-330, lesson 2)
Product feed
A file listing every product’s facts in fixed fields.
Taught in Feeds for AI shopping (COM-330, lesson 4)
Product group
The family that ties a product’s variants together.
Taught in Identifiers and variants (COM-330, lesson 6)
Product token
The bot's short name inside its user agent string, such as GPTBot.
Taught in AI crawlers and fetchers (GEO-300, lesson 1)
Prompt
What you type to tell the assistant what you want.
Taught in What an assistant really does (AX-100, lesson 1)
Prompt cluster
Prompts with the same underlying need that one page can answer.
Taught in From prompt set to content map (AEO-210, lesson 2)
Prompt injection
Hidden text that tries to give the agent new orders.
Taught in Guardrails that hold (AX-300, lesson 8)
Prompt library
A shared list of prompts that worked, ready to reuse.
Taught in Improve it in turns (AX-100, lesson 6)
Prompt panel
A fixed set of prompts asked the same way every period, so results can be compared over time.
Taught in Prompt sets that mirror buyers (GEO-310, lesson 2)
Prompt set
The list of real buyer questions you track and optimise for.
Taught in Prompts, not keywords (AE-200, lesson 10)
Prompt weight
A prompt's share of expected demand, used to weight the metrics built from it.
Taught in Prompt sets that mirror buyers (GEO-310, lesson 2)

Q

Query fan-out
An AI assistant splitting one question into several smaller searches, then combining what it finds.
Taught in From ten blue links to one answer (AE-200, lesson 1)
Query string
The part of a URL after the question mark, where utm_source travels; redirects can drop it.
Taught in AI referral plumbing (GEO-300, lesson 11)
Quoted passage
The exact part of your page an assistant uses in its answer.
Taught in Measuring content in AI search (AEO-210, lesson 12)

R

Range
A number reported with its interval, such as 30 to 46 bookings.
Taught in Reporting to leadership (GEO-310, lesson 12)
Raw-HTML test
Fetch a page as the bot, with JavaScript off, and read what comes back.
Taught in What fetchers see (GEO-300, lesson 3)
Read-only tool
A tool that can look things up but change nothing.
Taught in Give it the right tools (AX-300, lesson 4)
Referral test
A scripted click from an AI link that checks the source survives to the final page.
Taught in AI referral plumbing (GEO-300, lesson 11)
Referrer
The page a visitor came from, sent by the browser under the Referrer-Policy.
Taught in AI referral plumbing (GEO-300, lesson 11), Tracking AI referrals (GEO-310, lesson 6)
Regression set
Cases you re-run after every change to catch new breaks.
Taught in Test it with evals (AX-300, lesson 7)
Repeat work
A job you do again and again in much the same way.
Taught in Spot the repeat work (AX-200, lesson 1)
Repeat-work log
A week’s record of each repeat job, with its minutes.
Taught in Spot the repeat work (AX-200, lesson 1)
Repurposing
Turning one piece of content into formats for other channels.
Taught in Distribution for AI visibility (AEO-210, lesson 11)
Reputation prompt set
The questions people ask about the brand by name, asked the same way at every audit.
Taught in The reputation audit (REP-320, lesson 3)
Resolution page
What happened, what changed, when, and who verified it.
Taught in When old crises linger (REP-320, lesson 10)
Response fit
Matching the response to how responsible people hold you.
Taught in When old crises linger (REP-320, lesson 10)
Retire rule
When you’ll switch a workflow off, decided in advance.
Taught in Measure and improve (AX-200, lesson 10)
Retrieval
Finding and ranking the passages that best match a question.
Taught in How AI reads your content (AEO-210, lesson 1), Choose what it reads (AX-300, lesson 6)
Retrieval-augmented generation (RAG)
Searching for relevant passages first, then writing an answer grounded in them.
Taught in How an AI answer is made (AE-200, lesson 2)
Return rate
The share of orders that come back.
Taught in Returns and disputes (COM-330, lesson 11)
Reverse DNS check
Look up the IP's host name, then confirm that name points back to the IP.
Taught in AI crawlers and fetchers (GEO-300, lesson 1)
Reversible action
One you can fully undo, such as editing a draft.
Taught in A person in the loop (AX-300, lesson 9)
Review gating
Asking only happy customers for reviews, or steering unhappy ones away.
Taught in Honest review programs (REP-320, lesson 7)
Review response
A public reply to a review with the facts, the fix and the date.
Taught in Honest review programs (REP-320, lesson 7)
Risk score
What a mistake costs, and how easily it slips through.
Taught in Score it before you build (AX-200, lesson 3)
robots.txt
A file at the root of your site that tells crawlers which paths they may fetch.
Taught in Can AI load you? (AE-200, lesson 4)
Rollback
Going back to the last version that passed its evals.
Taught in Improve it safely (AX-300, lesson 12)
Rule
A fixed if-then test that decides where work goes.
Taught in Chain the steps (AX-200, lesson 6)
Run
One asking of one prompt in one assistant at one time; the smallest unit you log.
Taught in Sampling AI answers (GEO-310, lesson 4)

S

sameAs
A schema.org property that links your site to your official profiles as the same entity.
Taught in Your digital footprint (AE-200, lesson 7), Structured data as a graph (GEO-300, lesson 4)
Sanity check
A quick rough estimate that catches a number that’s off.
Taught in Numbers and spreadsheets (AX-100, lesson 11)
Scope
The exact access a connection grants, such as read-only.
Taught in Connect with MCP (AX-300, lesson 5)
Sectioning element
article, section, nav and aside: the HTML elements that say what a block is.
Taught in Semantic HTML for passages (GEO-300, lesson 5)
Self-contained passage
A section that names its subject and answers fully without needing the rest of the page.
Taught in Answer-shaped content (AE-200, lesson 6), Stand-alone passages (AEO-210, lesson 5)
Self-reported attribution
Asking customers how they found you, to catch what referrers miss.
Taught in From prompt to page to visit (GEO-310, lesson 7)
Server-side rendering
Sending the finished HTML from the server, so content is readable without running JavaScript.
Taught in Can AI load you? (AE-200, lesson 4), What fetchers see (GEO-300, lesson 3)
Severity
Reach (how many prompts and assistants carry a claim) times harm.
Taught in The reputation audit (REP-320, lesson 3)
Share of answer
How often your brand is named in AI answers for a defined set of prompts, compared with competitors.
Taught in SEO, GEO and AEO: what carries over (AE-200, lesson 3)
Share of voice (AI)
Your share of all brand mentions across the answers in the panel, against named competitors.
Taught in The metric dictionary (GEO-310, lesson 3), Benchmarking competitors (GEO-310, lesson 8)
Shipping details
Delivery cost and time, written as labelled data.
Taught in Facts on the page (COM-330, lesson 5)
Shopping agent
AI that compares products and can buy for a person.
Taught in How an agent picks (COM-330, lesson 1)
Sign-off
Your final read, as the reader, before it goes out.
Taught in Check before you send (AX-100, lesson 7)
Signed agent
An agent that proves who runs it with a digital signature.
Taught in Let trusted agents in (COM-330, lesson 9)
Signed request
The agent signs each request with a key your CDN can check.
Taught in AI crawlers and fetchers (GEO-300, lesson 1)
Soft 404
A page that says "not found" but returns 200, so bots keep fetching it.
Taught in What fetchers see (GEO-300, lesson 3)
Source citation
A link to where a fact comes from.
Taught in Evidence and specifics (AEO-210, lesson 7)
Source error
A wrong claim that traces to a real source; fix the source.
Taught in Fakes, errors and rivals (REP-320, lesson 11)
Source family
A kind of source AI draws on, such as reviews or news.
Taught in Where AI gets its opinion (REP-320, lesson 1)
Source line
The sentence and page a summary point came from.
Taught in Research and summaries (AX-100, lesson 10)
Source list
The short list of sources the agent may search.
Taught in Choose what it reads (AX-300, lesson 6)
Source mix
The blend of source families behind one answer.
Taught in Where AI gets its opinion (REP-320, lesson 1)
Source of truth
The one place a fact is stored; pages and APIs both read from it.
Taught in APIs and MCP for brands (GEO-300, lesson 8)
Specifics
Concrete details: numbers, dates, places, names.
Taught in Stand-alone passages (AEO-210, lesson 5)
Spoofed request
A request with a real bot token from an address the vendor does not publish.
Taught in AI bot logs, verified (GEO-300, lesson 10)
Stale copy
An old version of a fact still published somewhere.
Taught in When facts disagree (COM-330, lesson 3)
Stale source
A source whose facts were true once but are now out of date.
Taught in Where AI gets its opinion (REP-320, lesson 1)
Standing instructions
The rules it follows in every chat, written once.
Taught in Build a reusable assistant (AX-200, lesson 4)
Step
One small job in a workflow, such as filling the card.
Taught in Chain the steps (AX-200, lesson 6)
Step budget
The most steps, time or money one run may use.
Taught in Write the job card (AX-300, lesson 2)
Store-wide policy
One return or shipping policy that covers every product.
Taught in Facts on the page (COM-330, lesson 5)
Stratified prompt set
Prompts grouped by intent and segment, so every group buyers belong to is represented.
Taught in Prompt sets that mirror buyers (GEO-310, lesson 2)
Structured data
Machine-readable labels (schema.org) that describe the entities and facts on a page.
Taught in Speak machine (AE-200, lesson 5)
Structured formatting
Lists, tables and headings that show how facts relate.
Taught in Stand-alone passages (AEO-210, lesson 5)
Structured output
An answer split into named fields in a fixed order.
Taught in Templates and structured outputs (AX-200, lesson 5)
Supporting sentence
The exact line in a source that backs the claim.
Taught in Open the source (AX-100, lesson 8)
Syndication
Republishing content on partner sites with a link to the original.
Taught in Distribution for AI visibility (AEO-210, lesson 11)

T

Targeted summary
A summary built around the questions you need answered.
Taught in Research and summaries (AX-100, lesson 10)
Template
A fixed layout that every output must follow.
Taught in Templates and structured outputs (AX-200, lesson 5)
Test set
Real past cases, run again after every change.
Taught in Test before you trust (AX-200, lesson 8)
The loop
Audit, fix, publish, measure, double down, then repeat.
Taught in The answer loop (AE-200, lesson 12)
Third-party source
Any site or profile not controlled by you that talks about your brand.
Taught in Many voices (AE-200, lesson 8)
Three-check routine
Spot the claims, source them, sign off as the reader.
Taught in Check before you send (AX-100, lesson 7)
Threshold alert
A rule that flags a metric when it crosses a line you set in advance.
Taught in Dashboards that drive action (GEO-310, lesson 10)
Time to first byte
How long a fetcher waits before the first byte of the response arrives.
Taught in Fast, reliable responses (GEO-300, lesson 9)
Token
A small piece of text, often part of a word.
Taught in What an assistant really does (AX-100, lesson 1)
Tone
The feeling an answer carries about the brand: positive, mixed or negative.
Taught in Facts and feelings (REP-320, lesson 2)
Tool
An MCP function an assistant can call, with a schema and a description.
Taught in APIs and MCP for brands (GEO-300, lesson 8), What makes an agent (AX-300, lesson 1)
Tool description
Plain words telling the agent what a tool does, and when.
Taught in Give it the right tools (AX-300, lesson 4)
Trace
The step-by-step record of one agent run.
Taught in Watch every run (AX-300, lesson 10)
Training cutoff
The date after which a model has no built-in knowledge unless it retrieves live information.
Taught in How an AI answer is made (AE-200, lesson 2)
Training data
The text a model learned from before you used it.
Taught in Where answers come from (AX-100, lesson 2)
Training setting
The switch that decides if your chats train models.
Taught in Keep private data out (AX-100, lesson 9)
Transcript
The words of a video or audio file, published as text.
Taught in Beyond text (AEO-210, lesson 9)
Trigger
A measured threshold that moves an issue up the ladder.
Taught in Watch, escalate, respond (REP-320, lesson 12), Chain the steps (AX-200, lesson 6)
Trusted source
The original record, or an expert body you can rely on.
Taught in Check before you send (AX-100, lesson 7)

U

Unwritten rule
A rule people follow that nobody has written down.
Taught in Map the steps (AX-200, lesson 2)
Update note
A dated correction added to the original page or article, so the old source tells the new truth.
Taught in Fix it at the source (REP-320, lesson 4)
User agent
The name a crawler sends with each request, used in robots.txt rules and server logs.
Taught in Can AI load you? (AE-200, lesson 4), Reading crawler logs (GEO-310, lesson 5)
User-triggered fetch
A request made because a person asked an assistant something right now, such as ChatGPT-User or Claude-User.
Taught in Reading crawler logs (GEO-310, lesson 5)
User-triggered fetcher
A bot that fetches a page because a person asked an assistant right now.
Taught in AI crawlers and fetchers (GEO-300, lesson 1)

V

Variant
One version of a product, such as a size or colour.
Taught in Identifiers and variants (COM-330, lesson 6)
Verification pipeline
The steps that turn a log line into a verified bot, its job and a verdict.
Taught in AI bot logs, verified (GEO-300, lesson 10)
Verified-bot rule
A CDN or WAF rule that acts on a verified bot category, not on a user agent string.
Taught in Bot policy as code (GEO-300, lesson 2)

W

Winning prompt
In Sophyx, a prompt whose published pages actually earned AI-crawler reads, indexing and visits.
Taught in Measure what wins (AE-200, lesson 11), Winning prompts (GEO-310, lesson 11)
Workflow map
Each step of a job, in order, as it is done today.
Taught in Map the steps (AX-200, lesson 2)
Working memory
What the agent can see in this run; it ends with the run.
Taught in Plans and memory (AX-300, lesson 3)
Worth score
Hours saved a month, minus time to build and run it.
Taught in Score it before you build (AX-200, lesson 3)

Learn the terms in context

Answer Engine Specialist explains these ideas in twelve short lessons, with a certificate at the end. Regular price USD 89. Your first certification is free.

Free with any paid Sophyx plan. Certificates take up to 30% off your plan. See plans

Book a Demo