Product
Introducing aeoh Agents
Turn one Customer Prompt into a 30-day learning loop with daily AI visibility observations, weekly evidence reviews, and clear opportunities for what to improve next.
Published on
July 28, 2026
Written by
Maciej Czypek
Founder
Today we are introducing aeoh Agents: focused workers that keep one part of an AI Recommendation Funnel moving after the first plan is complete.
The first is the Customer Prompt Visibility Agent. Give it one buying question and it spends 30 days observing how that question is answered, turning repeated evidence into a clearer view of who appears, which sources matter, and what the business can improve next.
Why an Agent, not another score?
One AI answer is useful, but it is not a market. Recommendations can change between otherwise identical requests. Search may be used in one answer and not the next. The businesses, order, and sources can move too.
A single check hides that variation. A broad visibility score hides the buying journey. The Agent keeps the unit of work small and useful: one Customer Prompt, one business, and a repeated observation protocol.
One Agent follows one buying journey
A Customer Prompt represents a specific decision a customer might ask AI to help make. It might be finding a specialist in a city, comparing providers for a use case, or choosing a service under particular constraints.
Keeping the Agent attached to that prompt makes every result easier to interpret. The observations, competitors, cited sources, weekly reports, and recommended work all belong to the same customer decision.
What the Agent observes
01
Three observations a day
The Agent asks the same natural Customer Prompt three times on each scheduled day. Closely spaced observations make disagreement visible instead of allowing one answer to stand in for the whole market.
02
Recommendation evidence
Each valid observation records whether the business appeared, where it ranked, which other businesses were recommended, whether web search was used, and the answer itself.
03
Source evidence
When the answer uses the web, the Agent keeps consulted sources separate from sources actually cited in the answer. That distinction makes the next content or outreach decision more defensible.
Web search is available to the model, but the Agent does not force it. That is deliberate. We want to observe whether the system chooses to search for the prompt, then keep the answer and source trail it actually produced.
Weekly direction, not daily noise
Individual observations remain available, but the product is organized around the Customer Prompt. Every seven days, the Agent reviews the valid observations from that window and turns them into a report.
A competitive view of the Customer Prompt
The report summarizes valid observations into an appearance rate and competitive rank, then shows which businesses and sources recur. Missing technical checks are excluded rather than counted as a negative result.
An Intent Page opportunity
If competitor-owned content repeatedly appears in cited evidence, the Agent suggests an original page angle adapted to the business and the Customer Prompt. It does not recommend copying a competitor page.
A third-party source opportunity
If an independent source repeatedly appears, the Agent can suggest a legitimate editorial or outreach opportunity. The recommendation is grounded in observed recurrence, not a generic publisher list.
Four weekly reviews show how the prompt develops during the run. A final 30-day review closes the period without deleting the underlying evidence.
Built for honest measurement
The Agent uses a standardized OpenAI observation protocol with ChatGPT-related models. It is designed to make checks comparable over time. It is not a promise to reproduce every person's ChatGPT account, memory, location, interface, or model routing.
We also separate absence from failure. If a valid answer does not include the business, that is a real negative observation. If a technical check is missing or partial, it stays in the history but is excluded from visibility calculations.
Recommendations are evidence-backed opportunities, not causal claims. A cited page can be associated with an appearance without proving that it caused the appearance. The job of the Agent is to make the next experiment better informed.
A history that keeps getting more useful
The 30-day period ends. The evidence does not. Observations and reports remain attached to the Customer Prompt, so a business can return to the full history later or run another 30 days and compare periods.
That also makes Agents useful for businesses with several buying journeys. Launch one for the prompt that matters most, or launch separate Agents for different services, locations, and customer decisions. Each stays focused on its own evidence.
The first of more aeoh Agents
Customer Prompt Visibility is the first Agent because it closes an important loop: plan how a business can become recommendable, observe what AI actually does, then use the evidence to decide what to publish or pursue next.
Future Agents can take on other focused jobs inside the Funnel. The rule will stay the same: a clear scope, evidence the customer can inspect, and a useful outcome without pretending automation removes uncertainty.
Available now
Launch an Agent for one Customer Prompt
A 30-day Customer Prompt Visibility Agent is a one-time $79 purchase. It includes three scheduled observations a day, weekly evidence reviews, a final report, and permanent history. There is no subscription or automatic renewal.
Build a Funnel, choose the Customer Prompt that matters, and launch its Agent.
Build your Funnel