Three layers of GEO spending
GEO money goes to three layers; buying the wrong one is a more common mistake than overpaying. Sort the layers first; the numbers follow.
Tools: subscription pricing — you are buying measurement. A monitoring tool runs a fixed panel of buyer questions across AI engines regularly and reports whether your brand is mentioned, which sources get cited, your share of voice against competitors, and whether engines describe you correctly. Peec's public pricing page (checked 13 November 2026) lists three monthly tiers — Starter $95, Pro $245, Advanced $495 — differing mainly in prompts, projects and multi-country coverage; enterprise custom-quoted. Profound publishes a free trial and a custom enterprise plan — no fixed monthly price listed. Either way, price moves with engines, markets, prompts and seats. A tool tells you whether you are mentioned; it does not get you mentioned.
Diagnostics: fixed price, fixed deliverables. A diagnostic turns audit data into a prioritised action list, not another score report. Negency's package is a fixed RMB 19,800 (published on the diagnostic package page): human review, competitor comparison, 90-day roadmap and walkthrough call, with no per-page or per-finding surcharges.
Execution: quoted by scope and markets — it cannot be standardised. Machine-readable website work, question-shaped content, third-party mentions and periodic sampling reviews: starting points and target markets differ, so the same named service can mean several times more work for one company than another. A total price quoted without looking at your current state is not worth taking seriously.
Why execution has no standard price
Four factors decide the quote, and missing any one makes the number meaningless: starting point — the weaker the technical surface, the more engineering first; markets and languages — one market and five are different jobs; depth of mention building — directories, press and community each cost differently; re-measurement cadence — monthly costs more than quarterly.
How far apart can starting points be? The Negency Flagship Report 2026 (Issue 1) — 30 Chinese exporters' global website home domains; Negency AI visibility audit engine, around 30 signals, five categories; single-day sample, 30 October 2026; the report measures website technical visibility only (whether AI can crawl and read the site), not brand mention rates in AI answers — puts the median at 80/100, the mean at 80.1, the range at 55–98: 7 sites above 90, 6 below 70. The same "website remediation" means a few detail fixes at 90 points and a rebuild of the machine-readable layer at 55 — 17 of 30 sites had no JSON-LD at all, 15 of 30 had sitemap problems, 13 of 30 had a missing or non-unique H1. When the workload differs by multiples, the price cannot be the same.
Take two companies: A scores above 90 technically, so its bottleneck is mentions and spend goes to media and directory work. B scores below 60, machine readability well under the 69.2 sample mean, and needs structured-data and entity-fact repair first. Both buy "GEO services"; the work is entirely different. That is why a credible provider measures before quoting: without current-state data, any total price lacks a basis.
What should exist before any quote
Verifiable current-state data, in three parts: an audit score with itemised findings, a baseline mention rate on a fixed question set, and key competitors' mention rates. Without those three, any total price lacks a basis. Negency's sequence: free audit (machine score) → RMB 19,800 diagnostic (human review + 90-day roadmap) → engagement quoted from the roadmap. Every step can stop where it is; none binds you to the next.
The free audit turns "where does our technical surface stand" into a verifiable number, so pricing starts from shared facts. To see the diagnostic deliverable first, read the diagnostic sample excerpt.
Red flags
Beware two kinds of pricing: fees for "guaranteed recommendations" in a named engine (nobody can guarantee that — Google's docs confirm there are no additional technical requirements for AI Overviews or AI Mode, and organic answer citations are not sold), and lump sums with no itemised deliverables. The first sells luck; the second sells an information gap.
Two more to watch: annual-only contracts with full prepayment and no pilot acceptance criteria — all risk lands on the client; fees per "ranking achieved" — billing against engine sampling nobody controls. Before choosing a provider, run the 12 due-diligence questions: padding in a quote shows up on the spot.