LinkedIn
A clean dashboard can show 847 prompts where your client is missing. It cannot write the fix or put it live on the client site. If the vendor leaves that part to your agency, you are buying monitoring, not execution.
That distinction gets buried under words like visibility score, white-label, auto-optimisation and GEO. Each sounds as though work is getting done. Each can also describe a report that noticed a problem and handed it back to you. If you sell that report as an AEO retainer, your team absorbs the fulfillment hours when citations do not move.
I wrote this glossary for agency owners who need to know what they are buying before they price what they are selling. Read it in the order you would meet the terms while building a retainer. Then keep the questions handy for the next demo.
AEO
Plain definition: AEO means structuring content so an answer engine can lift a clean, correct answer from it.
Think short answer blocks, clear headings, tables and schema that remove guesswork. A useful distinction is AEO as the extraction layer and GEO as the authority layer, both built on SEO rather than replacing it. Getting a page ready to supply an answer is one job. Getting your brand named as the trusted source is another.
Vendors stretch AEO to cover everything from rank tracking to content writing. A tracked gap is not an AEO change. Ask to see one client page before and after the vendor changed it, then ask where that page was cited. If the answer is only a score, you have found the boundary of the product.
GEO
Plain definition: GEO means building brand presence so answer engines name you when it matters.
That work can include citations, third-party mentions, consistent entity data, reviews and coverage on sites the engines draw from. The boundary with AEO matters more than the acronym: a vendor can make your own page easier to extract without doing anything to build authority beyond your domain.
One research breakdown priced AEO at $160.9M and GEO at $1,089.3M for 2026. A 6.8x gap shows how much rides on where someone draws that boundary. Do not let a broad label stand in for a scope of work. Ask which engines the GEO work covers and what the vendor publishes or fixes off your domain to earn a mention.
AI SEO
Plain definition: AI SEO means using AI to do SEO work faster.
That is all it means until a vendor tells you which work it does. When an agency searches AI SEO white label, it can find three different products:
- A content writer that drafts posts your team must edit and publish.
- A rank and prompt tracker that puts your logo on a PDF.
- An execution platform that writes, fixes and publishes under your brand.
I used to treat any AI help as progress. I was wrong. Resell the first or second product as the third and you keep the fulfillment work, plus the blame when nothing moves. Ask whose hours put the fix live after the software flags a gap.
Prompt tracking
Plain definition: Prompt tracking checks what answer engines say when buyers ask questions that could lead to a sale.
Not rankings. Answers. A useful setup runs a fixed list of buyer prompts across ChatGPT, Gemini and Perplexity and saves the responses, so you can see who was named and how. HubSpot’s AEO approach gives the baseline: which prompts mention you, how you are described and where competitors appear instead.
The sales-deck trick is to sell the number of prompts tracked as coverage. Five hundred trivia questions nobody buying from your client would ask are easy to count and hard to turn into revenue. Buyer intent matters more than prompt volume. Ask to see the 20 revenue prompts for your client type and how many answers cite the client today.
Citation share vs mention rate
Plain definition: Citation share counts how often you are a cited source in tracked answers; mention rate counts how often your name appears in their text.
Vendors mix the two because mention rate tends to look bigger. A name in an answer is worth knowing about, but it does not tell you whether the engine used your client as a source. For a retainer sold on citations, that difference is the deliverable.
A tracking roundup notes that share-of-voice tracking can flag a competitor pulling ahead in AI answers before pipeline reflects it. That warning is useful only when you can inspect buyer prompts by engine, rather than blending linked citations with stray brand echoes. Ask for cited share on your money prompts, split between ChatGPT and Perplexity, and ask which competitor took the citations you lost.
Answer mapping
Plain definition: Answer mapping assigns each buyer prompt you track to the page that should supply its answer.
One practitioner breakdown puts the sequence well: identify the answer that would satisfy the intent, then decide whether to update a page or create one. The prompt list alone is not the map. A tracker with 200 prompts and no page owner has given your agency a longer to-do list, not coverage.
Every priority prompt needs a destination and an owner. Ask for the URL mapped to each of your top 20 prompts, which answers have no page yet and who will write those pages.
Content execution vs content briefs
Plain definition: A content brief tells your team what to write; content execution gets the page live.
Many AEO tools stop at an outline with entities to include and headings to use. Your writers, editors and client approvals handle the rest. That handoff is where agency margin disappears if you sold an outcome and budgeted for a button click.
An agency tool roundup describes page-level audits as turning a visibility score into a to-do list. Fair enough, if you price the to-do list as a to-do list. Execution needs a publish step and an action log: pages drafted, schema added, internal links made and changes shipped. Ask what goes live after the brief without your staff touching it, and ask to see the log.
Knowledge graph management
Plain definition: Knowledge graph management keeps the facts answer engines find about your client consistent across the places they look.
MarTech describes two layers: what a model knows about the industry and where your client sits within it. A stale Google profile, directory listings with old phone numbers and a homepage that never plainly says what the firm does make that second layer harder than it needs to be.
The same article reports that 85% of AI-answer mentions in its analysis came from third-party sites, not brand domains. Editing one about page will not cover that territory. Ask which third-party sources the vendor fixes or earns, and how it checks whether answers change afterward. The work is unglamorous directory and schema work; our knowledge graph and client CMS walkthrough gets into the workflow.
CMS auto-publishing
Plain definition: CMS auto-publishing puts new or updated pages on the client site without your team copying them across by hand.
The honest version connects to the client’s CMS, prepares the page mapped to a prompt, then publishes or stages it for approval. The stretched version is an export button labeled publishing. An agency tool roundup pairs direct CMS auto-publishing with an AI visibility score. The pairing matters: publishing without measurement is blind, while measurement without publishing is theater.
Make the vendor show the handoff. Ask for a page its system published to a live client CMS, who approved it and how long approval took. If the process ends with someone on your team logging into 30 WordPress installs every Friday, put that labor in the retainer price.
Edge proxy
Plain definition: An edge proxy sits in front of a client site so changes can be served without editing its underlying CMS or theme.
It can be a way to deploy answer blocks, schema, internal links and technical SEO changes when the client’s developer queue moves slowly. Picture a law firm whose web person takes three weeks to approve a plugin. The proxy can let you ship a page-structure change sooner and roll it back if needed.
The misuse is selling the setup as the result. Setup is plumbing; execution is what flows through it. Ask how many pages and fixes the vendor shipped per client per month, then ask what happens to those changes if you cancel. If the second answer is foggy, the delivery method may become a hostage situation.
White-label
Plain definition: White-label means your client sees your agency rather than the vendor fulfilling the work.
Reports carry your logo, the sender domain is yours and vendor support stays out of the client relationship. For agencies, white-label delivery also means branded reports, quiet fulfillment and pricing you can resell. A logo upload field is not the whole service.
The common stretch is a PDF you can export and decorate in Canva. Your team still does the work and fields the 9 p.m. “why did our citations drop?” call without an action log to consult. Ask whether you can forward last week’s report untouched. Then ask whether any login or email exposes the vendor, and whether its strategist ever joins a client call. For the pricing shift behind that expectation, see From White-Label SEO to AEO.
Per-domain flat pricing
Plain definition: Per-domain flat pricing charges one fee for each client domain rather than separately metering every prompt, seat or page.
The phrase matters when you have multiple retainers to price. The pricing roundup lists Semrush’s AI Toolkit at $99 per month per domain billed annually for 25 daily tracked prompts, with additional charges for more prompts and seats. It lists HubSpot AEO at $50 per month for 25 prompts. Across 30 clients, the limits and add-ons affect your margin before you write a word.
The misuse is calling a domain fee flat while prompts, seats and domains each have their own meter. Ask for the invoice, not the adjective: if you double tracked prompts and add five domains, what exactly will you pay?
Prompt and keyword caps
Plain definition: Prompt and keyword caps limit how many buyer questions or search terms a plan will track or work on.
A 25-prompt cap sounds tidy until a plumber needs service prompts across three engines, plus questions that name competitors. Then you are deciding which buyer questions to leave out. Vendors call caps focus. For an agency, they can mean rationing.
The nastier version offers uncapped tracking but capped execution: you can watch 500 prompts decay while the plan fixes ten. Find out which activity hits the ceiling. On prompt 26, does the vendor stop tracking, stop fixing or charge more? Ask to see the answer in the plan. I break down the scope question further in Flat-Rate AEO Pricing Per Domain.
Done-for-you
Plain definition: Done-for-you means you sell the retainer while the vendor fulfills the agreed work under your agency’s name.
For an AEO retainer, that could mean tracking prompts, mapping pages, writing content, shipping technical fixes, working on citations and sending a branded report. The stretched version is a dashboard plus a strategy call where someone tells your junior what to do.
Put hours beside the label. Say you charge $2,000 per month per client and fulfillment takes six hours at $75 an hour. That leaves $1,550 per client before other costs or revision rounds. It is not free margin, and the number shrinks as soon as the vendor hands more work back to you. Ask for live URLs, dates and action logs showing what shipped without agency staff.
groas for agencies is built around that test: execution runs around the clock, actions are logged with their reasoning and weekly reports go out under your brand. On any vendor demo, I would still ask to see the log. That is how you find where its product ends and your labor begins.
Visibility score
Plain definition: A visibility score blends the share of tracked prompts where your client appears into one number.
HubSpot’s version runs across ChatGPT, Gemini and Perplexity, with prompt-level detail underneath. That makes a score useful as a trend line. It does not make the score a deliverable. A climb from 12 to 18 can look good on a slide while the client still loses the prompts that lead to quote requests. The gain may sit entirely in long-tail questions. What the deck calls blended visibility, I call a rate hiding its denominator.
To sort monitoring from execution, I run a demo through three columns:
- What gets tracked?
- What gets shipped to a client domain?
- What shows that the answer changed?
Monitoring tools fill the first column and stop. One tool roundup found only one tool that auto-corrected listings instead of merely reporting them. Ask for the action log, the published URL and the before-and-after answer on the same buyer prompt. Then ask which three prompts moved the score and why.
If I could retire one term tomorrow, it would be visibility score. Keep it in the dashboard if you like. Sell the work shipped and the buyer prompts where your client became the cited answer. Make the rep answer in those terms, not in theirs.