In GEO/AEO job postings from May-July, most postings did NOT mention a monitoring tool by name. But of those that did, 80.4% named Profound. Thirty-seven of ALL the 152 GEO/AEO job postings mention Profound. That’s 24.3%, about one in four.
It’s early, but there’s a clear, winning preference now.
Ten years ago, I wrote the most bookmarked thing I’ve ever published: “5 Questions to Ask Your Next Marketo Hire.” It resonated, because the people doing the hiring weren’t Marketo experts. Most of them weren’t marketing ops experts either. They knew they needed somebody who could run the system, but had no way to tell a real operator from somebody who had sat next to one for a year.
That’s where GEO is standing right now. And I’m in a unique position—having been an early Marketo user—I can see the similarities. I have been using Profound for a year now and can tell an expert from a poser.
Unlike Marketo, Profound has a very shallow floor. Get access Monday. One hundred prompts by Wednesday. Screenshot the visibility dashboard next Monday. In a portfolio review, that looks like competence, and if you’re a VP of Marketing who has never personally logged into an AI visibility platform, you have no way to see the difference between good and great.
Therefore, BEHOLD! Five interview questions for people hiring someone and expecting them to run Profound for them.
Question 1: How do you create prompts and organize them into reporting categories?
You’re testing for two things. A repeatable method for generating prompts, and a reporting structure they’ve decided on before the first export. Prompt architecture can’t cheaply be redone six months from now. This is critical.
Excellent answers will…
Start by naming a methodology. We use buying lenses, which I wrote about in Dentists Don’t Prompt ChatGPT Like Marketers Do, because buyer psychology matters. There are a dozen defensible frameworks and I don’t, personally, think it matters which one they bring. You mostly want one to exist at all.
You’ll also want to hear a brand-exclusion rule. Profound automatically creates competitive and sentiment prompts. So you’ll want to listen for the candidate that understands not to waste a series of prompts doing what Profound does out-of-the-box. This also has an impact on reporting. Once your brand is named in a prompt, the visibility score turns into a self-fulfilling prophecy. OF COURSE you’re visible when someone uses your name. Or good god, I hope so! Candidates who raise this on their own have likely watched a topic get contaminated and had to explain the numbers to somebody afterward.
They should also use the platform’s actual vocabulary and use it correctly. Topics and tags are different mechanisms doing different jobs, and a strong candidate will explain how topics roll up into the out-of-the-box dashboard reporting while tags give them a second axis to slice the same prompt set.
Agents coming up unprompted (yes, pun intended!) is a good sign, too. It means they’ve been beyond the reporting layer.
And they’ll describe the process as collaborative while still clearly owning the output. Input from sales and CS on a draft they built. Not a draft that sales built for them.
Meh answers will…
Hand most of the prompt creation to a committee. Gather input from sales, product marketing, and execs, then assemble the list from what comes back.
This isn’t wrong. Sales hears things you don’t, and cross-functional input is useful. It’s just not the level of ownership I expect from somebody who’s going to operate autonomously, because the committee will not be in the room in month four when the topic structure needs to change.
These candidates still use the right nouns. Topics, tags, visibility score. They almost never mention agents.
Bad answers will…
Talk around prompt creation entirely. You know the catchalls. “It’ll be custom to our use case.” “I’d listen to sales calls to hear what customers are actually saying.”
Those sound right, and that’s the whole problem. It’s been the correct-sounding answer on LinkedIn for eighteen months, so repeating it proves somebody has been reading, not necessarily DOING.
Follow-up: Make them do it live. Provide 10 prompts and ask them what they’d do with it.
Question 2: What do you think of Profound’s content creation?
The right answer here is probably more my taste and opinion than the rest, but again, you want to hire someone with OPINIONS, not just pleasantries.
Excellent answers will…
Hate it. Every person worth their salt in GEO that I’ve talked to has the same reaction to the content recommendations, and it usually arrives fast and unprompted. Profound is genuinely great at measurement and mediocre at ideation and content generation. Somebody with real hours in the platform has already tried the content features, gotten generic output, and gone back to writing.
Bonus points if they mention how adding a boatload of assets to the knowledgebase doesn’t make it any better.
Meh answers will…
Land somewhere measured. They’ll say things like, “It can be good and it can be bad, useful as a starting point for volume, needs heavy rewriting.”
Follow-up: Ask for one specific suggestion it made that they thought was good. Somebody who has actually used it produces an example in about four seconds, with a caveat attached. Somebody fibbing produces a category of suggestion instead of a suggestion.
Bad answers
Only praise. They’ll tout how this is one of the platform’s strongest features, going from visibility gap to content brief in one tool is a huge workflow win. A tool that does more than report, it moves to action! And so on. That’s the sales deck.
People who value these recommendations are usually following Profound’s content rather than working in the system.
Follow-up: Same ask, plus one more. What did they publish out of it, and what happened to visibility on those prompts afterward? The silence is the only answer you’ll need.
Question 3: What’s the most undercover Profound feature that you really like?
Here’s the truth about how most people use this platform: They create the prompts, organize them into topics and tags, screenshot the dashboard elements, paste into a deck, repeat next month. This question tells you how seriously a person took their management of Profound.
Excellent answers
Agent analytics, watched pages, the Platforms tab inside Answer Engine Insights, or knowledge bases.
Nobody stumbles into watched pages. These features take curiosity to find and a specific reason to care, and the reason usually shows up as a question the candidate couldn’t answer with the dashboard.
Meh answers
Something off the visibility dashboard, the content creation agents, or “agents” as a broad category.
Agents in Profound are challenging to build from scratch, which makes the agent answer cheap to test.
Follow-up: Have them walk you through one they created, start to finish. If SQL never comes up, they either took one off the shelf or the agent wasn’t doing much.
Bad answers
Sidestep the question, or quote the blog. Whatever Profound shipped most recently is the tell, and right now that means AIM and some version of “agentic measurement is where the category is heading.”
That’s a paraphrase of a launch post, and it correlates more reliably than anything else on this list with never having logged in.
Question 4: Which report do you use to first see the symptoms of an issue?
The best answer to this one rejects the premise, because most problems worth catching never surface in the Profound dashboard at all.
Excellent answers
Reframe the question. The dashboard gives you week over week, 14 days over 14, or month over month, and those windows are fine for spotting a spike. They cannot see a SLOPE.
So the answer I want involves routinely getting data OUT of Profound, via export or the API, and running trend analysis on it with Claude or Gemini or ChatGPT. Then going back into the dashboard to confirm the mechanism once they know what they’re looking for.
Do not ding a candidate who exports manually instead of using the API. Getting API access provisioned is a slog, Profound’s customer success is rough here, and even we struggled to get it. Manual export plus real trend analysis beats API access plus screenshots every single time.
Meh answers
Name a specific report on the visibility dashboard and then explain how they dig in from there. Topic performance shifts, expand to prompt level to isolate the problem prompts, open individual answers to see where the citations went. Usually with an example attached, like a review site displacing a client in three of four engines on comparison prompts.
That’s competent. There’s a drill path and there’s a mechanism instead of a vibe, and it’s the honest median of this field. What’s missing is any sense that the default time windows are a constraint and counterproductive.
Bad answers
Stop at the dashboard at a single level, as though the visibility score trend by itself will tell them something. It’ll give you a hint. The real answer is always two layers deeper, and a candidate who quits at layer one will bring you findings that are actually just symptoms.
Question 5: Explain how Profound’s visibility score is calculated.
This is a critical metric for GEO, partly because on its surface it gets misunderstood constantly. If this person is running GEO for you, they need to know WTF it actually is, because it’s the number that ends up in front of your CFO.
Excellent answers
Explain the mechanics in their own words, roughly: Profound takes every prompt and runs it through each supported engine every day. At minimum that’s three answers per prompt, from ChatGPT, Perplexity, and Google AI Overviews. Then it looks for recommendations of you or your competitors. If no brand is mentioned in an answer, that answer isn’t factored into the score at all. Only where some brand shows up do you get credit for being named or dinged for being absent.
The tell is whether they understand the DENOMINATOR. It’s “answers where a brand appeared,” not “answers we tracked.” Somebody who gets that also understands why the score can climb without their own performance changing, which means they won’t walk into a QBR and take credit for the models simply starting to make recommendations in a category.
Meh answers
Quote the definition accurately from Profound’s site without REALLY understanding it. If you’re hiring an expert, this isn’t a definition that should be quoted instead of contextualized. “The percentage of AI-generated responses that include your brand, divided by the total number of relevant responses tracked for your chosen prompts” is the literal Profound definition.
Plenty of working practitioners live right here and it’s not disqualifying. But somebody who can only recite the definition can’t explain an anomaly, and this metric produces anomalies constantly.
Follow-up: Ask what counts as “relevant.”
Bad answers
Something like “it’s how often you show up in AI answers.”
Hire that and your headline GEO metric gets reported as a percentage of everything, your CFO hears market share, and you spend a quarter walking back a number nobody understood in the first place.
What these five questions are actually measuring
Read them back and they collapse into one question. Are you in the tool, or are you following the tool?
Every excellent answer above shares a fingerprint. It uses the platform’s real vocabulary correctly, it volunteers a limitation no one expresses publicly, and it describes a workaround somebody found in a fit of frustration. That combination is close to impossible to fake from reading, which is exactly why it’s worth screening for.
Give the meh answers a fairer hearing than they usually get, though. Somebody who lands in the middle on all five has done real work inside a reporting layer and hasn’t had the autonomy or the client pressure to push past it yet. Whether that’s a problem depends entirely on what you’re hiring them into. Of the 37 postings that named Profound last quarter, 24 were manager, senior IC, or senior manager level. Translated: one person owning the whole function, with nobody above them who knows the platform well enough to catch a bad call.
So, three things to do with this:
- Put one of these in the phone screen, not the final round. Question 5 is the cheapest to grade and the hardest to fake.
- Ask every follow-up, including on the answers you liked. The first answer tells you what they’ve read. The follow-up tells you what they’ve done.
- Decide before the loop starts whether you’re hiring an owner or a ramp. Both are good investments. Paying owner comp for a ramp is not.
And if you’re reading this as the candidate instead of the hiring manager, good news: every excellent answer here is available to you this month. Build a prompt set on a framework you can defend out loud. Export your data and go looking for a slope. Go find watched pages. Thirty-seven employers named this platform last quarter and almost none of them have anybody in-house who could grade these five answers, which means the differentiation is sitting there unclaimed. That’s how the Marketo people got rich, and it didn’t stay unclaimed for long.
Frequently Asked Questions
What should you look for when hiring someone to run Profound? Look for a candidate who owns the prompt methodology, understands the visibility score’s denominator, and can describe features they’ve actually used. Somebody who only recites dashboard terminology without building anything beyond the reporting layer is a weaker hire.
How is Profound’s visibility score calculated? Profound runs every tracked prompt through each supported AI engine and measures how often a brand appears, but the denominator is answers where any brand appeared, not every prompt tracked. Candidates who misunderstand this distinction tend to report visibility as market share, which misleads stakeholders.
What is the difference between an owner and a ramp in this hiring context? An owner independently builds prompt methodology, exports data for trend analysis, and creates agents from scratch, while a ramp still needs guidance on these tasks. Both can be reasonable hires, but paying owner-level compensation for a ramp-level skill set is a costly mismatch.
Why do meh answers still get an interview pass? Candidates who land in the middle on all five questions typically have real hands-on experience with the platform, using correct vocabulary and mechanisms rather than vibes. They may lack full ownership, but that’s different from the vague, catchall answers that signal no real usage at all. And let’s be real, finding EXPERTS right now is challenging. This is all new.
