Inside the Head of AEO Role: The Real Workload
Here's how I hear most AI visibility programs start these days.
An executive comes back from a conference or reads a report while on a flight. Then they walk into the Monday meeting and say some version of:
“We need to be showing up in AI answers when our customers ask questions related to our product or service. Marketing: figure it out and report back.”
But the marketing person (or team) is already carrying 20 pounds in their 10-pound bags.
They already run the blog, the newsletter, the webinar program, and half of demand generation.
No additional headcount is offered to help get this new initiative off the ground; there's zero budget, and a check-in is scheduled for thirty days out. (Cue the panic sweat).
I want to walk through what that person has actually been handed, because I don’t think most people realize what’s being asked of them, or what a successful AI visibility program actually requires.
The ask to "get us cited by ChatGPT" sounds like a small add-on or to-do list adjustment at first glance, but so far, from my hands-on experience, it’s actually more like three distinct full-time jobs (if not more).
I already made the case for why I think we’re about to see a new role emerge around Answer Engine Optimization (AEO) ownership in a previous issue, but this is the version that breaks down the actual workload.
So if your boss tells you to figure out how to get the brand cited by AI tools, here’s a quick overview of what that’s going to take.
AEO job one: Measurement (+ why one check tells you nothing)
You’ve gotta start somewhere with this work and get a baseline for where you currently stand within the context of AI visibility.
Most people start by asking ChatGPT a few questions about their category, screenshotting the results, and pasting them into a spreadsheet.
But that's not measurement, my friends. Measurement of AI visibility is actually quite tricky.
The reason: None of the data for measuring this is static, and it’s all shaped by user context.
That’s why the AI answer your CEO gets on their phone and the one your marketing manager gets on a work laptop are two different tests of two different things (and why the answers are probably wildly different.)
Here's a shortlist of some contributing factors to why AI visibility is so variable:
Which engine you check. Each engine spits out different answers based on its own logic re: how it decides what to cite as a source. (Reminder: the decision-making logic for each AI engine is still a black box, and none of them have disclosed the theory or process behind what makes one source get cited over another.)
User context. Memory, chat history, location, and logged-in status all appear to shape what comes back as an answer. The tool wants to give the user the best possible answer based on everything it knows about him/her, so that context is going to shape its answer (unless you’re in an incognito chat or logged out.)
Chat history. Follow-up questions inherit context from earlier in the thread, so the same prompt asked cold and asked mid-conversation aren't the same prompt.
How the question is worded. Small phrasing changes can result in very different answers.
With all this in mind, think about what it means for your Monday screenshot: You caught one roll of the dice.
Which brings me to one thing I want to hammer on: AI visibility is largely a probability.
I will do a whole issue on the probabilistic nature of AI answers at some point, but the short version is that this isn't something you rank for the way you did with SEO.
Measuring AI visibility means running many prompts, many times, across multiple engines, in clean sessions with no history or personalization, and then watching how that trends over time.
So…what are the AI visibility tools actually telling you?
There are a lot of AI software tools available in 2026 (Profound, AirOps, Peec AI, Scrunch, AthenaHQ, OtterlyAI, Ahrefs Brand Radar, Semrush's AI toolkit, and more launching monthly).
They’re mostly gauging a similar handful of metrics:
Presence rate. Across your prompt set (AKA the questions your audience members ask), how often do you show up? This is the headline number most dashboards lead with.
Share of voice. When you and your competitors are both eligible for a question, who gets named? This is the number your executive will likely care about, because it's comparative.
Cited vs. mentioned vs. recommended. Cited means a link back to your page, mentioned means your name appears in the answer text, and recommended means the AI actively suggests you as an option.
Which sources are getting pulled. Not just whether you appeared in an answer, but which URLs the engine reached for when it answered.
Sentiment and accuracy. How you're being described, and whether the model is saying things about you that are outdated or incorrect.
Prompt-level breakdowns. Which specific questions you win and lose, by engine.
One big thing that still hasn’t been solved yet: None of these tools tell you WHY you were cited or not cited.
Instead, they're largely observation systems. They'll show you that you lost ground on a prompt, but figuring out what changed and what to do about it is a human job (and one that can easily take up 40 hours a week).
This is just job one, folks. Let’s move on to job two.
AEO job two: The prompt management piece
Your measurement is only as good as the questions you're tracking. These are called your “prompts,” and they are essentially the questions your prospective customers ask an AI tool when researching which product or service to buy.
Most teams build their prompt list in about twenty minutes, using the questions they wish buyers were asking. (Or worse: they buy an AI monitoring software tool and use the prompts it generates automatically based on the limited context it has about the company during the initial onboarding sequence. Don’t do this!!!)
So what are the questions you think your prospective buyers are asking? Something like "What's the best [category] platform?" may feel important and obvious, but this is probably not how real buyers phrase things.
Real potential customers will ask questions that are much more specific and based on their use case/the problems they need to solve.
Realistic prompts look more like:
How hard is it to migrate from [competitor] to [your company]?
What's the difference between [approach A] and [approach B]?
Is [category tool] worth it for a [type of company] team of 20?
What do people complain about with [tool]?
Building a realistic set of prompts is a research project, not a brainstorm. And it needs to happen before you start measuring your progress, because it’s the foundation of your whole strategy.
That means looking for REAL customer language you have via:
Sales call recordings
Support tickets
Product reviews/review sites
Community threads on forums like Reddit
Your own keyword/SEO data on what’s driving new traffic
BUT THAT’S NOT ALL, FOLKS! From there, you have prompt maintenance. Your category vocabulary shifts, competitors launch and rename things, new use cases show up…you name it. The prompts you track need to evolve alongside your product or service, and with your customers' language, too.
Translation: A prompt set built in January could need a major revamp by June, so somebody has to be generating and retiring prompts on an ongoing basis, and tracking which ones changed so your trend lines still mean something. This is a ongoing job. Not a setup task.
AEO job three: The off-site piece of the puzzle
Now the biggest piece. Your own website accounts for something like 5 to 10% of what AI models pull from when answering questions about you, per McKinsey.
Which means the overwhelming majority of the opportunity is on surfaces you don't control. According to Search Engine Land, platforms like LinkedIn, Reddit, and YouTube are some of the most-cited in AI answers (at least right now).
That means an effective AI visibility program needs to stand up initiatives across those surfaces. Each one is its own mini-program and requires input from various in-house teams. The work includes things like:
Employee experts publishing under their own names on a platform like LinkedIn. Finding them, screening them, getting their time formally allocated, giving them editorial support, keeping them consistent. (When LinkedIn content gets cited, 59% of it comes from individual profiles rather than company pages, which is why this one is usually first.)
Customer voices and review platforms. Getting detailed, specific reviews written on the review sites that get pulled into comparison answers.
Community presence. Reddit, niche forums, Slack and Discord groups, with genuine participation rather than drive-by promotion or astroturfing.
Earned media and podcasts. Slower, sometimes more expensive, but still valuable.
Managing external creator partnerships, once the internal layer is working.
Notice that every single one of those requires other people to do things.
That's the part that makes this a full-time role rather than a task. The AEO owner isn't producing all this content themselves; they're recruiting participants, scheduling, coaching, unblocking, chasing, and reporting.
This is my main objection re: telling marketing "just figure it out.”
A marketing manager with four other responsibilities can maybe handle the measurement piece, but they can’t also stand up five internal programs that depend on colleagues in other departments who don't report to them (and have their own priorities.)
Painting a clear picture about what AEO work entails
When your boss says, “Get us cited by ChatGPT,” this is the picture you need to paint about what this work actually entails:
Measurement and monitoring: a day or two a week once tooling is in place, often more during setup. From there, it’s running prompt sets and monitoring weekly performance to see what’s working and what’s not.
Prompt set work: This is a heavy up-front time commitment (call reviews, ticket mining, competitive research), then a steady few hours weekly for generation and maintenance.
Off-site content efforts: This work will require lots of cross-team coordination and thinking beyond the company website. Recruiting and supporting employees alone will eat a day a week once you have more than a handful of participants.
These three bullets alone could build a 40-hour workweek. What’s more: A Head of AEO is a role with a strange shape: part analyst, part program manager, part editor, part internal evangelist. Am I seeing more new AEO roles pop up every week on my AI search jobs board? Yes, I am. The roles are starting to pop up, and more companies are beginning to dedicate salaried positions to oversee this work.
Read the full AI search hiring report with all the details on this emerging Head of AEO role.
If you can't hire for a Head of AEO yet, do this
It’s still early, there are a lot of moving parts to this work, and as quickly as things change in this space, some companies are waiting to see how things shake out.
However: 1 in 3 searches start with AI. This is a rapid consumer behavior shift, and it’s not going away. You can sit on the sidelines and see how things shake out, but it’s going to get more competitive over time (not less.)
My advice is: start small, but start now.
Pick one engine and a small prompt set. Twenty well-researched prompts on one engine (say, ChatGPT), run consistently, will be more helpful than two hundred prompts run once across four different engines. Consistency and data trends over time is what makes the intel useful.
Get a baseline before you change anything. Run your set, record where you stand, then don't touch it for 90 days. Without a real “before”, you can’t demonstrate an “after”, and demonstrating an after is how you make a case for extra budget + headcount around AI visibility efforts.
Start off-site work with employees. Your people know your product best, and their expertise shared across LinkedIn helps kickstart this work.
Protect the time explicitly. If this is going to become 30% of someone's job, take 30% of something else off their plate and have it documented by their manager. Now that you know what a big ask this is, if someone’s getting tasked with AI visibility, their workload likely needs to be adjusted.
Two closing thoughts.
Tooling helps with the measurement piece, but it isn't free. Most of the measurement burden I laid out can be automated with a visibility platform. Candidly, I’ve heard some marketing teams are buying software to measure AEO as a way to appease AI-enthusiastic leadership and say, “Yep, we’ve got a tool for that in place,”...but they aren’t really sure what to do with it yet.
This new role is still taking shape. I'm describing what the work requires based on what I see running these programs, so there’s still some delegation that needs to happen over the next 18 months before we know what a Head of AEO’s job description looks like. (P.S. If your program looks different and it's working, I'd like to hear about it!)
In short: AEO and “get us cited by ChatGPT” requires a multi-pronged approach.
It’s not easy, it’s not quick, and that’s why right now I’d recommend hiring it out to either an agency partner or IC before tasking your already maxed-out in-house marketing team with it.
FAQ about the workload required for effective AEO
Is AEO prompt research a one-time project? No. A prompt set requires ongoing maintenance, because the language around it keeps moving. Category vocabulary shifts, competitors launch and rename products, and new use cases appear. A prompt set built in January can need a serious revamp by June. Somebody has to be generating and retiring prompts on a continuing basis, and documenting which ones changed and when, or your trend lines stop meaning anything. Treat prompt management as a standing responsibility rather than a setup task you complete once.
Do I need to buy an AI visibility tool? A visibility platform will automate most of the measurement burden, and that is real time saved. What no tool on the market currently does is tell you why you were or weren't cited. These platforms are observation systems: they will show you that you lost ground on a prompt, but diagnosing what changed and deciding what to do about it remains a human job. Buying software and treating the initiative as handled is a common move right now, and it doesn't work. The tool answers what happened. Somebody still has to answer why, and then go fix it.
What do AI visibility tools measure? Most AI visibility tools track a similar set of metrics. Presence rate is how often you appear across your prompt set. Share of voice is whether you or a competitor gets named when both are eligible for a question, and it's usually the number executives care about because it's comparative. The tools also separate being cited, which means a link back to your page, from being mentioned in the answer text, from being actively recommended as an option. Beyond that, expect reporting on which URLs the engine pulled from, sentiment and factual accuracy in how you're described, and prompt-level wins and losses broken out by engine.
If most AEO work isn't on my website, where is it? Most of the AEO opportunity sits off-site. Your own website accounts for something like 5 to 10 percent of what AI models pull from when answering questions about you, per McKinsey, which means the overwhelming majority of the surface area belongs to platforms you don't control. According to Search Engine Land, LinkedIn, Reddit, and YouTube are among the most-cited sources in AI answers right now, alongside review platforms, podcasts, and niche communities. Each of those is its own mini-program, and nearly all of them depend on colleagues in other departments agreeing to participate.
Why start off-site AEO work with employees? Employee publishing is the off-site layer you have the most influence over, and it's the one with the clearest payoff. When LinkedIn content gets cited in AI answers, 59 percent of it comes from individual profiles rather than company pages. Your subject matter experts already understand the product better than any agency will. The work is finding them, screening them, getting their time formally allocated by their managers, giving them editorial support, and keeping them consistent over months rather than weeks. None of that is content production, which is exactly why it takes a person.
How much time does an AEO program take? An AEO program adds up to roughly a full-time role. Measurement and monitoring run a day or two a week once tooling is in place, and more during setup. Prompt set work is a heavy up-front commitment of call reviews, ticket mining, and competitive research, then a few steady hours weekly for generation and maintenance. Off-site coordination will consume about a day a week once you have more than a handful of employee participants. That's a forty-hour week before anyone writes a word, which is why this doesn't fit on top of an existing job without something coming off the plate.
We can't hire a Head of AEO yet. What should we do? Start small and start now. Pick one engine and twenty well-researched prompts rather than trying to cover everything at once. Get a baseline and leave it alone for 90 days, because without a real before you can't demonstrate an after, and demonstrating an after is how you make the case for budget and headcount later. Begin your off-site work with employees, since they're the participants closest to you. And protect the time explicitly: if AI visibility is going to become 30 percent of someone's job, take 30 percent of something else off it, in writing, with their manager's sign-off.
Can we wait and see how AI search shakes out? You can wait, but the position gets harder rather than easier. Roughly 1 in 3 searches now start with AI, which is a consumer behavior shift rather than a trend cycle, and the competitive field will be more crowded whenever you decide to enter it. Companies that spend the next year building a baseline, a prompt set, and an employee publishing habit will have compounding data that late entrants won't. If the internal capacity genuinely isn't there today, the more realistic near-term move is an agency partner or a contractor rather than handing it to an already maxed-out in-house team.
