AEO for Agencies: How to Sell AI Visibility and Audit Clients at Scale

AEO fits because it extends work your team already does: crawlability, accurate facts, useful pages and third-party mentions. Google says SEO best practices remain relevant for AI Overviews and AI Mode. Sell AEO as SEO fundamentals plus a new way to measure them; clients who have read Google's documentation will spot hype. Answer engine optimization (AEO), also called generative engine optimization (GEO), is the practice of improving visibility in AI-generated answers.

In AEOeye's recommendation benchmark (127 audits, 892 answers, September 2026), the audited brand was named in 157 answers (17.6%) and absent from the other 735. Answers named 2.9 competitors on average. When a brand was ranked, it came first in 71.4% of 147 cases, so being named at all is the hard part (aggregate data).

Treat that as a pattern in our records, not a market rate: the brands were self-selected, and engine samples are unequal and not comparable.

Sell three products in sequence: a fixed-price baseline audit, a fix sprint and a re-test. Each ends in a decision the client can accept or decline, which makes the offer easy to buy and to scope.

Baseline audit. Use a documented checklist such as our AEO audit checklist. Record mentions, citations, competitors and any factual errors per answer.

Fix sprint. Work in this order: access, facts, pages, corroboration. OpenAI says sites that disallow OAI-SearchBot will not be shown in ChatGPT search answers, so check crawler access before you write anything.

Re-test. This is the only layer that earns a retainer. Without it the client cannot tell whether the sprint changed anything.

Check whether AI retrieval bots can reach the site at all. Most brands do not block them, so a quick look rules the problem in or out.

In AEOeye's crawler access study (October 2026, 384 sites, a convenience sample skewed to established brands), 1 of 193 SaaS sites blocked any AI retrieval bot in robots.txt. For news publishers it was 27 of 38. The study's per-site data is public.

Stores need a second check. In the same study, 24 of 88 e-commerce homepages (27.3%) challenged or blocked our self-identified crawler, which robots.txt does not show. Ask the developer to check CDN and firewall logs for OAI-SearchBot, Claude-SearchBot and PerplexityBot. Limits: one fetch per site, one day, main host only.

Standardize the collection and keep the diagnosis human. A senior should spend time deciding which gap matters, not copying answers out of browser tabs.

1. Write a question template per vertical. Use unbranded buyer questions such as "best [category] for [use case]" and "alternatives to [competitor]". People phrase the same need very differently: in SparkToro's survey, 142 prompts about choosing headphones had a semantic similarity of 0.081. Yet the same four brands appeared in 55-77% of 994 responses.
2. Freeze the protocol. Record engine, model tier, date, locale and runs per prompt. The same study found under a 1-in-100 chance of identical brand lists across 2,961 runs, so repeat each prompt.
3. Screen prospects cheaply. AEOeye's free audit asks ChatGPT three unbranded buyer questions and shows the full answers and the competitors named. It is a quick pitch screen, not a baseline: one engine only.
4. Baseline each signed client. The $29 one-time full report adds Claude, Gemini, Google AI and Perplexity, and 10 questions instead of 3. Each engine runs at the default tier a free user sees.
5. Buy a tool for ongoing monitoring. AEOeye has no agency plan, white-label output or multi-client dashboard, and it is not a daily tracker.

For daily monitoring across a roster, use a tracker built for it. As of October 2026, Peec AI's agency plans list Essential at $245 a month (10,000 credits), Growth at $495 (25,000 credits, "3-10 clients") and Scale at $795 (65,000 credits, "11-25+ clients"), plus free pitch projects for prospects. A credit is one prompt, one model, one day (credits documentation). By our arithmetic, Growth works out to $49.50 to $165 per client per month at its stated client range. This is documentation-based; we have not used Peec hands-on.

Price the diagnosis and the shipped fixes, not the tool. Tool cost per client is small, so cost-plus pricing underprices your judgment and invites the question of why the client is not buying the tool directly.

Market anchors are thin and come from sellers. As of October 2026, Rock The Rankings lists strategy programs from $4,500 a month at its 90-day rate ($500 a month less than month-to-month), a 30-day AI Search Blueprint from $10,000 one-time, and a one-hour session at $500. That is one agency's price list, not an industry rate.

Our position: fix the price of audits and sprints; let the retainer pay for re-tests and continuing fixes. Quote from these drivers:

• Audit: brands, competitors, prompts, engines, locales and runs per prompt, plus the readout.
• Sprint: number and type of pages, developer dependencies, review rounds and outreach.
• Re-test: cadence, sample size and reporting depth.

Write exclusions and change-order triggers into the proposal, and never sell guaranteed mentions: nobody controls what a model writes on a given run. On the tool side, AEOeye's pricing is a free audit plus the $29 one-time report per brand, with no subscription.

Report mention rate across repeated runs on a frozen prompt set, next to named competitors and cited sources. Skip any single "AI ranking position".

SparkToro's author concluded that visibility percentage across many prompts, run multiple times, is a reasonable metric, while any tool giving a ranking position in AI is, in his words, "full of baloney". The study involved Gumshoe.ai, an AI tracking vendor, which its author disclosed.

A client scoreboard that holds up:

• Mention rate: share of sampled answers that name the client.
• Competitor share: who else appears on the same samples.
• Cited sources: URLs engines link, reported as observed.
• Accuracy: whether the answer states the client's price, features and category correctly.

Add Google's data where it applies. Google's generative AI performance reports in Search Console show impressions, pages, countries, devices and dates for AI Overviews and AI Mode, and reached all websites by August 31, 2026. They are not prompt-level and the announcement lists no click metric, so pair them with your own runs (metric definitions).

Keep negative results in. In our benchmark, 144 of 157 mentions were positive: the usual problem is invisibility, not bad press.

Position on method, not outcomes: name your vertical, show your protocol and put a re-test in the contract. "We rerun the same buyer questions after every fix sprint" is believable. "We will get you into ChatGPT" is not.

Audit your own agency first and publish what you change. When we checked our own site in September, Search Console showed about half of 496 posts were unrelated to AI visibility, with 0 clicks in 90 days. We retired 234, and on October 3, 2026 merged another 165 pages into 42. One strong page per question beats five thin ones, and that is a story you can tell clients.

Publish client case studies only with written permission and real before-and-after runs. Until then, sell the rigor of the method.