Somebody at your company has been made responsible for AI search. In most companies it’s the SEO lead, and the reasoning took about four seconds. It has “search” in the name. Search is that person’s job. Done.
It’s the most natural assignment on the org chart and it’s the reason a lot of these programs sit still for two quarters. Not because the person is bad at it. Because you handed the outcome to the one role that’s structurally unable to pull the main lever.
The work is mostly not on your website
Here’s the part that breaks the assignment. When an engine answers a question about your category, the answer is assembled from what other people have published about you. Review pages, comparison posts, analyst write-ups, community threads, practitioner blogs. Your own site is the weakest input in that pile, because a machine trained on how careful humans read the web has learned to discount the vendor describing itself.
So the actual work splits roughly like this. A minority of it is on your domain: be technically readable, be clear about who you are, publish proof that’s specific enough to lift. That part is genuinely search-shaped and the SEO lead is the right person for it. The majority of it is off your domain: get a stale comparison page corrected, get an analyst to use the right noun for your category, get a credible third party to describe the segment you actually win, get a real customer to say it somewhere findable.
Look at that second list and ask a plain question. Who at your company can make any of those things happen?
Not the SEO lead. They can publish, optimize, measure, and report. They can’t commission an analyst note. They can’t get a review site to update a two-year-old page. They usually can’t email a customer without going through customer marketing, and they usually have no standing relationship with the trade press. Their tools, their access, and their mandate all stop at the domain boundary, and the lever they need is on the other side of it.
Then, at the end of the quarter, they get asked why the AI search numbers haven’t moved.
It’s an authority problem, not a skill problem
This is the part that gets misdiagnosed. When the program stalls, the instinct is to conclude the owner didn’t know enough about AI search, so the company sends them to a conference, buys a monitoring tool, or hires a consultant to teach them the tactics.
The tactics were never the blocker. Watch where these programs actually stop and it’s always the same place: the step that requires somebody else to do something. The outreach that needs comms to approve the angle. The customer quote that needs the CSM to make an introduction. The category language that needs product marketing to agree to stop inventing new words for what you sell. Each of those is a small ask, and each one sits in a queue behind that team’s own roadmap, because the person asking has no claim on their time.
So the useful question isn’t “who understands AI search best.” It’s “who at this company can get an outside party to change what they publish about us, and get internal teams to help them do it.” That’s a comms and product marketing question. In most B2B companies the honest answer is product marketing, sometimes comms, occasionally a founder who still has the relationships. Whoever it is, they need a standing claim on PR, customer marketing, and content, not a request form.
The SEO lead doesn’t disappear in that arrangement. They keep the on-domain work, which is real, and they keep the measurement layer, which nobody else can run. Knowing which sources the engine is actually pulling from is a research skill that looks a lot like search work, and it’s what tells the owner where to aim. That’s a strong second seat. It’s a weak first one.
The two ways companies get around this, and why both fail
The first is to give it to nobody in particular. AI search becomes a shared priority, which means it becomes a line in four different plans and a real commitment in none of them. Everyone does the piece that overlaps with work they were already doing, which is always the on-site piece, because that’s the piece that requires no cooperation. Six months later the site is clean and the answer hasn’t changed.
The second is to create a new owner from scratch. Head of AI Search, or the AI search pod. This feels decisive and it usually reproduces the original problem with a better title, because the new role arrives with a mandate and no authority. They can convene meetings. They can’t make PR pick up their outreach ahead of the launch that PR is already judged on. A new owner without a claim on other teams’ capacity is the same stalled program in different packaging.
What works is less exciting. Put the work with someone who already has cross-functional standing, give them the SEO lead as the permanent research and measurement partner, and make one specific team’s cooperation a written expectation rather than a favor. That’s an org design decision, not a marketing tactic, which is why it usually has to be made a level above whoever has been quietly assigned this so far.
What I got wrong
I used to recommend the obvious thing. Name a single owner, put them in SEO because that’s where search lives, hand them the playbook, get out of the way.
I did exactly that on an engagement and watched it stall for a quarter. The owner was good. She executed everything inside her control quickly and well: the site got clear, the proof got specific, the technical floor got fixed. Then she hit the source layer and stopped, because every remaining move needed somebody else. The stale G2 profile needed customer marketing to run a review push. The analyst who had them filed under the wrong category needed a real relationship to reopen. The comparison page with the wrong pricing on it needed an outreach email nobody had authorized her to send. She raised all of it. It went into queues. Nothing came out.
I read the flat stretch as normal at first, because the early flat stretch usually is normal, and that cost us several more weeks before I looked past the calendar and at the blocked list. The problem wasn’t patience. It was that the entire remaining plan required cooperation she had no way to demand.
What fixed it was moving ownership to their product marketing lead, who already sat in the room with comms and customer success every week, and keeping the SEO lead on as the measurement and source-research half of the pair. Nothing about the tactics changed. The same plan, the same list of stale sources, the same target pages. It started moving within weeks, because the asks were now coming from someone whose asks got answered.
I had been treating this as a question of who knows the most. It’s a question of who can get other people to act.
How to make the call at your company
Take the list of things that would actually change what the engine says about you. Not the content calendar. The specific list: which third-party pages are wrong, which sources the engine leans on for your category, which credible outsider has never described you correctly.
Then, for each one, write down the name of the person who can make it happen and what they would need from someone else. If most of the rows require cooperation from a team the current owner can’t direct, you don’t have an AI search execution problem. You have an ownership problem, and no amount of tactical education will resolve it.
Give the work to whoever can clear those rows, keep your search person on the research and the reporting, and tell your leadership what the program is actually being judged on before the first quarterly review, so the flat stretch doesn’t get read as failure by people watching a traffic chart.
If you want help mapping where your answers are being written, and figuring out who inside your company can realistically go change them, that’s the work we do and the method we use.