Ask five AI-visibility agencies to price a program and you will get five numbers with no common unit. The published market data now exists to cut through it: First Page Sage’s October 2026 report, built from 1,146 GEO/AEO price quotes collected from 214 agencies between July 6 and October 2, 2026, puts the median GEO/AEO retainer at $6,850 per month — up 66% from $4,120 in Q1 2024, and 45% above the $4,740 median for traditional SEO retainers (First Page Sage, “GEO & AEO Pricing Statistics: 2026 Report,” October 7, 2026: https://firstpagesage.com/seo-blog/geo-aeo-pricing-statistics/).
One disclosure before the numbers: First Page Sage is an agency with a paid program, so treat its survey as a well-documented market sample from a participant, not a neutral census. The methodology is stated on the page (published pricing pages, client proposals, signed contracts; all quotes normalized to monthly figures), and it is the largest GEO pricing dataset currently published — but the conflict of interest is real, and the rest of the market’s data is thinner.
This is the anatomy piece: the five components inside any AI-visibility retainer, what each costs in labor, why the published medians look the way they do, what the $500–$1,000 tier should actually include, the red flags that mean you are paying for theater, and the two buyer profiles for whom a retainer is the wrong product. It is the cost-side companion to AI visibility tools vs. managed service (the comparison from the other direction) and to the decision framework for whether to invest at all.
The five components
Every AI-visibility retainer, at any price, is some combination of five things. Here they are with what each one actually costs in labor:
1. Query definition and maintenance — the intellectual core
Choosing the money queries (category, comparison, alternatives, pricing, pain-point), structuring the query set per buyer type, and updating it as the market and the AI surfaces shift. This is the part that most differentiates a good program from a bad one, because it is pure judgment — buyer research plus category knowledge — and it cannot be automated meaningfully. In a strong retainer this is senior-level work. In most retainers it happens once at onboarding and quietly stops.
2. Scheduled runs — cheap tooling, real coordination
Running the query set across ChatGPT, Perplexity, Gemini and Claude on a schedule, capturing the answers verbatim, and handling the surface-specific mechanics (login-gated surfaces, UTM parameters, referrer behavior, per-surface rate limits). This is mostly engineering plus tooling. The raw cost is low — open hosted tooling plus per-call API costs runs to tens of dollars a month at this scale (our own stack runs on roughly $10/month of hosted tooling plus API cents, per our about page) — so when a retainer’s tooling bill is a visible line item at five figures, you are paying for margin on software, not for infrastructure.
3. Scoring and diffing — cheap if automated, expensive if manual
Turning captured answers into scores (cited / mentioned / absent, answer quality, ranking position within the answer) and diffing month over month to find what changed and why. This is the component that separates “we ran the queries” from “we can tell you what your changes did.” It is cheap when the pipeline exists and manual-and-expensive when each report is assembled by a human reading 40 answers.
4. Fixes — the only component that changes your visibility
Everything above changes your knowledge of your visibility. Only fixes change the visibility: structured data and content changes on your site, plus the off-site work that determines whether any AI surface cites you at all — the review-platform, community, comparison-media and catalog presence that the ChatGPT citation data shows dominates. This is where senior time actually goes, and it is the component to check for first in any proposal. If a retainer’s scope is components 1–3 with no fixes, you are buying a dashboard with a subscription.
5. Reporting and narrative — cheap to produce, expensive to fake
A short monthly read: what moved, what it means, what happens next. Five minutes of reading. The red-flag version is the 30-slide deck that reports impressions, impressions, impressions and then a “next steps” section containing no commitments — because the scope was measurement-only and there is nothing to do.
Why the published medians look the way they do
From the same First Page Sage report (October 2026, linked above), the tier data:
| Tier | Monthly range | Median | Typical minimum term | Avg. deliverables/month |
|---|---|---|---|---|
| Starter (local, small business) | $1,800–$4,200 | $2,950 | 3 months | 3.4 |
| Growth (mid-market) | $4,500–$11,000 | $7,400 | 6 months | 6.5 |
| Advanced (upper mid-market) | $11,000–$22,000 | $14,600 | 6 months | 9.8 |
| Enterprise (large enterprise) | $22,000–$60,000+ | $31,500 | 12 months | 14.2 |
Three facts from the report explain the shape, and map directly onto the five components:
- Output rises more slowly than price. Enterprise clients pay 10.7× the Starter median for 4.2× the monthly deliverables — because the higher tiers buy PR placements, multi-market coverage and senior staffing. In component terms: Starter is mostly 1, 2, 3 and 5 (a query set, some runs, a report), while the premium tiers are where component 4 — actual fixes, plus the off-site work — becomes real scope.
- The scope at the low end is stated plainly. Starter engagements “usually cover reporting, a technical audit, and a few pages each month.” Read that against the component list and the implication is explicit: at $2,950/month you are buying measurement plus a little content work, not a program that moves your visibility.
- The mid-market median is where “program” stops being a word. At the $6,850 all-tiers median, a serious program should include scheduled runs, diffing, and a defined number of fixes with named owners. If the deliverable list at that price is measurement-only, compare it against the Starter scope above — you are paying 2.3× for the same five components with better fonts.
The report also gives the in-house reference point: an all-in Growth-tier agency engagement costs $98,400 per year, versus $363,512 per year for a three-person in-house team covering the same function — which is the arithmetic behind “you do not build this in-house to save $100k, you build it in-house because you want the capability permanently.” (Same source; same agency-published caveat.)
For the pricing structure context: 68% of the 214 agencies price GEO as a monthly retainer (41% custom, 27% tiered), and only 23% of agencies publish any pricing — so most of the market’s numbers will never appear in any survey.
What a $500–$1,000/month program should look like
This is the tier the published data barely reaches (the Starter median is 3× higher), which is exactly why buyers cannot evaluate offers at this price — nobody anchors to anything. The honest contents, at this price:
- A defined money-query set of 10–20 queries across the four surfaces, structured by buyer type, and reviewed quarterly.
- Scheduled monthly runs on open-stack tooling (hosted infrastructure plus per-call API costs — the tooling itself is a rounding error at this scale, so it should not be priced as a line item).
- A one-page monthly report: per-surface citation scorecard, the diff from last month, and one or two findings. One page, not a deck.
- A bounded number of fixes — a small monthly commitment (site-side technical or content changes, and off-site citation work where applicable), with the list negotiated in, not discovered later.
- The query set, captures and raw scores as your property. If the data is not yours, the service is a subscription to someone else’s findings.
What it should not be: “AI visibility monitoring” at $500–$1,000 with no query list, no fix commitment and a login to a generic dashboard is a tool with a service wrapper — and a tool is what you can buy for a few hundred dollars per month. The line is not the price; it is whether a named human owns component 4.
Red flags: when you’re paying for theater
In order of how often we see them:
- Deliverables measured in “visibility” without a query list. No query list, no per-surface scope, no scoring rubric. If the engagement letter cannot name the queries being measured, the number in the report is unchallengeable — and unchallengeable is the product.
- A dashboard with no owner. You receive login credentials to a monitoring tool and a monthly note. That is a SaaS purchase billed as a service; the same tool standalone is a fraction of the retainer price.
- Measurement-only scope at mid-market prices. Components 1–3 and 5, no component 4, at a $6,000+ price. The Starter tier publishes its measurement-only scope at a third of the price.
- Unattributed statistics in the pitch. “AI search is X% of searches” or “brands cited earn 11× more” with no source, date or URL. If the vendor cannot cite the data, the data is for the pitch, not for your decision. (We label our own numbers the same way: see the methodology.)
- Terms that don’t match the tier. A 6–12 month lock-in at Starter-tier price means you are expected to cancel late, not that the service is that good. The published tier data shows terms rising with price (3 → 12 months) because senior work amortizes — a long term at the low end is a term, not a commitment.
- Tooling billed as a line item. When the infrastructure cost is tens of dollars a month, a “platform fee” is margin on software. Fine if disclosed; a red flag if padded.
The mirror-image flag: a vendor that never names a query, a surface or a rubric, even when asked. The five components above are the checklist — a straight answer about which ones are in scope, at what depth, by whom, is the entire due diligence.
Who should skip the retainer
Two profiles where the honest answer is “not yet”:
- An in-house SEO team that can run the pipeline. If someone on staff can maintain a query set, run the four surfaces on a schedule, and read a diff, the retainer’s components 1–3 are duplicative. Buy the tool (a few hundred dollars a month), run it, and spend agency money only on component 4 — the fixes that need senior judgment or off-site reach. That is the in-house-plus-tooling path, and it is the right one for teams whose SEO capability is already real.
- A budget under ~$1,000/month with no owner. A sub-$1,000 “program” cannot staff the fixes, and a measurement-only purchase is a tool wearing a service’s coat. At this budget, buy the tool or run the manual checks in our four-surface guide yourself — the information value is identical; only the ownership is missing, and ownership is what the money should buy next, not the measurement.
The decision framework — including when measurement alone is enough and when it isn’t — is in is AI search worth it for B2B SaaS.
See what the deliverables look like
Our own pricing is published on the about page, to the tier the market data doesn’t reach — and the free quick check produces the same per-surface snapshot (cited, mentioned, absent, with what each answer said) before you commit to anything. Get the free quick check.