How to Sell GEO Services: Package AI Visibility for Clients

To sell GEO services, sell measurement and a fix plan, not placement. Package them into tiers by how often you track: each tier costs a known number of tracking runs and can detect a known size of change. Never promise that ChatGPT will recommend a client, because no one controls that.

The cadence decides what a tier can prove. A monthly snapshot pins a 10 % mention rate to about ±5 points, weekly tracking to about ±2.4 and daily to about ±0.9. By the end of this post you'll have three tiers built on that maths, the cost of each at real credit prices, proposal wording that holds up, and a calculator for your own numbers.

Data used: Tier precision and costs come from the calculator below, using BugViso's AI Visibility Tracker prices (October 2026: $5 per run, or 30 runs a month for $139) and the shape of our own prompt set: 35 unbranded prompts answered by 4 engines per run. The example deliverables come from a real report on bugviso.com: 400 answers from 50 prompts × ChatGPT, Gemini, Perplexity and Claude × 2 runs on 2026-10-20, against five anonymised competitors.

What a GEO Service Actually Delivers

A GEO (generative engine optimisation) service tells a client how often AI engines recommend them on buyer questions, against which competitors, citing which sources, and what to change to be named more often. It's a measurement and advisory service, and the honest version says so.

Here's what one real report contained, for our own brand:

Notice what isn't in the table: a ranking position on an engine, a traffic forecast or a guarantee. Google's documentation on AI features and your website says there are no additional requirements or special optimisations for appearing in AI Overviews. Anyone selling a secret switch is selling the wrong thing.

Why You Can't Sell Placement

AI answers change from run to run, so a promise about any single answer is a promise you can't keep. Re-asking the same prompt minutes later changes which brands are named, as our study of AI answer consistency found. In our AI brand sentiment data, the label for the same brand on the same prompt and engine flipped on 31 % of repeated answers.

What does hold steady is the rate over many answers. Pooled mention rates barely moved between our two runs: the leader went from 35 to 36 mentions out of 140, and our brand from 11 to 10. So sell the rate, the trend and the work, and let the client see the answers behind them.

Eligibility is the only part with firm rules. For ChatGPT search, OpenAI's crawler documentation says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers. Checking crawler access is a legitimate onboarding deliverable. Promising what comes after it isn't.

How to Sell GEO Services in Three Tiers

Build tiers by tracking cadence, because cadence sets both your cost and the smallest change you can report as real.

The precision column is the 95 % Wilson interval half-width for a pooled unbranded mention rate of 10 %, from runs × 35 unbranded prompts × 4 engines. It's what you can call a real change in the monthly report. Our guide on how often to track AI visibility explains why the margin shrinks the way it does.

One caveat to say out loud: daily runs re-ask the same prompts, so they tighten the estimate for that prompt set but don't fix a prompt set that misses how buyers ask. Precision past ±2 points matters less than a good set of questions.

Set your fee from the cost floor and the hours, not from a market rate. There's no reliable public benchmark for GEO service pricing, and an invented one won't survive a client's second question.

What Goes in Each Tier

Every tier gets the same measurement; higher tiers add frequency and work, not different metrics. That keeps reports comparable when a client moves up.

The action plan is where the margin is. Measurement is cheap to run; deciding what to change and doing it is skilled work, as our guide to the AI visibility action plan shows. In our own report the top three actions targeted 250, 130 and 112 missed answers. Each is a content or outreach project in itself.

Write a GEO Proposal That Holds Up

A good proposal states the method, the scope and the limits before the price. Clients who understand the method don't churn when a single answer goes the wrong way.

Scope, in plain terms:

• The questions: 50 buyer prompts, drafted from the client's site, approved by the client and frozen for the contract term. Edits create a new version, as covered in prompt versioning.
• The engines: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, asked through their APIs with web search on.
• The competitors: at least three, named by the client.
• The cadence: monthly, weekly or daily runs, per the tier.

Limits, also in plain terms. Paste these into the proposal:

AI engines vary their answers from one run to the next. We report rates over many answers with intervals, and we only call a change real when it's outside that interval. We can't guarantee that any engine will name or recommend you. API answers can differ from what a signed-in user sees in an app.

Then the deliverables table from the tier you're selling, then the price. In that order.

Price a Tier for Your Own Numbers

This standard-library calculator turns brands and cadence into runs, credits, the cheaper of pay-as-you-go and the monthly plan, cost per brand, your margin at a fee you choose, and the precision a month of data buys. Defaults match our prompt set and October 2026 prices, so change --credit and the plan flags if prices change.

#!/usr/bin/env python3
"""geo_package_calc.py: price a GEO (AI visibility) service tier from runs, cost and precision.

Usage: python3 geo_package_calc.py --brands 5 --cadence weekly --fee 400
       [--credit 5] [--plan-price 139] [--plan-credits 30] [--included 0]
       [--unbranded 35] [--engines 4] [--rate 0.10]
"""
import argparse, math

RUNS = {"monthly": 1, "weekly": 52 / 12, "daily": 30}


def wilson_half(p, n, z=1.96):
    d = 1 + z * z / n
    return 100 * z * math.sqrt(p * (1 - p) / n + z * z / (4 * n * n)) / d


def cost(credits, a):
    payg = credits * a.credit
    top_up = max(0, credits - a.plan_credits) * a.credit
    plan = a.plan_price + top_up
    return (payg, "pay as you go") if payg <= plan else (plan, f"monthly plan + {max(0, credits - a.plan_credits)} credits")


def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--brands", type=int, default=1)
    ap.add_argument("--cadence", choices=list(RUNS) + ["all"], default="all")
    ap.add_argument("--fee", type=float, default=0, help="your monthly fee per client brand")
    ap.add_argument("--credit", type=float, default=5)
    ap.add_argument("--plan-price", type=float, default=139)
    ap.add_argument("--plan-credits", type=int, default=30)
    ap.add_argument("--included", type=int, default=0, help="runs a month your agency plan already includes")
    ap.add_argument("--unbranded", type=int, default=35, help="unbranded prompts per run")
    ap.add_argument("--engines", type=int, default=4, help="engines that return answers")
    ap.add_argument("--rate", type=float, default=0.10, help="typical mention rate to size precision for")
    a = ap.parse_args()
    cadences = list(RUNS) if a.cadence == "all" else [a.cadence]
    print(f"{a.brands} brand(s); {a.unbranded} unbranded prompts x {a.engines} engines per run; "
          f"precision sized for a {100 * a.rate:.0f}% mention rate\n")
    print(f"{'cadence':<9}{'runs/mo':>8}{'credits':>9}{'cost/mo':>10}  {'cheapest option':<32}"
          f"{'per brand':>10}{'margin':>9}{'± pts':>7}")
    for c in cadences:
        runs = RUNS[c]
        credits = max(0, math.ceil(runs * a.brands) - a.included)
        total, how = cost(credits, a)
        per = total / a.brands
        margin = f"{100 * (a.fee - per) / a.fee:.0f}%" if a.fee else "-"
        n = max(1, round(runs * a.unbranded * a.engines))
        print(f"{c:<9}{runs:>8.1f}{credits:>9}{total:>10.2f}  {how:<32}{per:>10.2f}{margin:>9}"
              f"{wilson_half(a.rate, n):>7.1f}")


if __name__ == "__main__":
    main()

Real output for five client brands, with an illustrative $300 fee and the 2 runs a month that the Agency Pro plan includes:

$ python3 geo_package_calc.py --brands 5 --fee 300 --included 2
5 brand(s); 35 unbranded prompts x 4 engines per run; precision sized for a 10% mention rate

cadence   runs/mo  credits   cost/mo  cheapest option                  per brand   margin  ± pts
monthly       1.0        3     15.00  pay as you go                         3.00      99%    5.0
weekly        4.3       20    100.00  pay as you go                        20.00      93%    2.4
daily        30.0      148    729.00  monthly plan + 118 credits          145.80      51%    0.9

The run cost is small next to the fee at weekly cadence. What the margin column leaves out is your time on the action plan, which is where most of the cost sits. Price the hours separately, and the tool cost stops being the question.

How BugViso Supports a GEO Service

BugViso's AI Visibility Tracker produces the measurement half of the service for each client brand, with its own prompts, competitors and schedule. It asks ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, all with web search on, a set of 50 buyer prompts it drafts from the client's site, which you review and edit. Each brand is tracked against at least three competitors you name.

Each run produces the deliverables in the first table: the visibility index and leaderboard, mention rates with intervals, the sources each engine cites and the gaps against competitors, a fact check of claims about the brand, and a ranked "what to do" plan with AI-written drafts and impact tracking. Runs go on demand or on a daily, weekly or monthly schedule, and reports export as PDF or CSV or go out as a shareable link. Credits cost $5 a run or 30 a month for $139, and agency plans include 1, 2 or 7 runs a month. The details are on the AI visibility tracker page.

Mistakes That Lose GEO Clients

• Selling rankings in ChatGPT. Answers vary by run. Sell rates, trends and the work.
• Quoting an industry benchmark. There's no shared prompt set behind one. Compare against named competitors, with the paired tests in our AI visibility benchmark guide.
• Reporting a score without the plan. A number with no next step reads as a cost, not a service.
• Editing prompts mid-contract. It breaks the trend line. Version the set and say so in the report.
• Selling daily tracking to every client. ±0.9 points is precision most clients can't act on. Weekly fits most retainers.
• Skipping the accuracy check. Wrong prices in AI answers cost deals quietly, and fixing them is a quick, visible win.

Frequently Asked Questions

What are GEO services?

GEO services measure and improve how often AI engines such as ChatGPT, Gemini, Perplexity and Claude recommend a brand on buyer questions. A typical service tracks a fixed prompt set against named competitors, checks what the engines claim about the brand, finds the sources they cite instead, and runs a prioritised plan of content and outreach fixes.

How much should I charge for GEO services?

Start from your cost floor and your hours, not from a market rate. At October 2026 prices, weekly tracking of one brand costs about $25 a month in runs and daily tracking $139. Most of the cost of a GEO retainer is the work on the action plan, so price that by the hours it takes. Avoid quoting industry price ranges, as none are reliably published.

Can an agency guarantee ChatGPT will recommend a client?

No. AI answers vary between runs, and no one controls which brands an engine names. What an agency can guarantee is the method: a fixed set of buyer questions, tracked on schedule against named competitors, with honest intervals and a plan of work. Crawler access (for example, allowing OAI-SearchBot) is the only hard eligibility requirement.

How often should GEO results be reported to clients?

Monthly reports on weekly data fit most retainers. Weekly tracking pins a 10 % mention rate to about ±2.4 points over a month, which is enough to show real movement. Daily tracking suits brands actively shipping fixes. A single monthly run (±5 points) works for snapshots and pitches, not for showing progress.

What should a GEO proposal include?

The prompt set and how it was approved, the engines, the named competitors, the tracking cadence, the deliverables per tier, and a plain limits paragraph saying answers vary and placement can't be guaranteed. Put the method and limits before the price, so the client buys the right expectations along with the service.

Sell the Method, Then the Work

The tier table is the product, the limits paragraph is the trust, and the action plan is the margin. Agencies that lead with those three can price GEO services on their own terms. Run the measurement side for each client with BugViso's AI visibility tracking.