AI Can Ship SEO at Scale, but It Still Can't Decide What to Rank For

Agentic systems have made it possible for a tiny team to produce the volume of SEO work that used to require a full agency. That's the easy part now. The hard part, deciding what a business should actually rank for, what to ignore, and when a reported win is just a vanity metric, still lands on a person. The operators getting real results out of these tools treat the automation as an engine and keep a human accountable for the strategy, not just the output.

Deana Rock and Al Levy are the Co-Founders of RevRocket, an SEO and AEO agency in St. Petersburg, Florida built for small businesses that have been priced out of serious search work. They designed the company as complements. Rock runs the technical engine, configuring agentic systems across the client base, deploying the work, and monitoring whether the results are real. Levy takes the founder's seat on margins, market position, and where a category is heading, then works backward to what the work should be. The pitch is both halves in continuous collaboration, at a fraction of what a traditional agency charges.

That split is the whole thesis. The automation moves fast; a person keeps it right. "There's a misconception that AI can do everything," says Rock. "It accelerates certain things, but the human in the loop is still very important." What the human supplies is the judgment the tools can't, and it starts with how the two of them decide what's even worth trying.

The engine and the strategist

RevRocket is deliberately a team of two, and Rock's read is that the technology is what makes that viable without falling behind it. The pace of change in agentic SEO and AEO is relentless, from prompting to vibe coding to loops to graphs, and she frames staying ahead of it as the daily job. "We're a team of two, and we feel empowered by the technology," says Rock. "We're on top of it right now, instead of it running our day." The point of that leverage is reach: serving small businesses that can't match a larger competitor's market spend, and doing it lean.

Deciding which of the endless new tactics to actually adopt comes down to method rather than hype. Levy describes an approach built on curiosity and a willingness to abandon what isn't working, testing a tactic long enough to know whether it holds up. "We dive into everything and work with it long enough to figure out whether it really has meat on the bone," says Levy, describing the nimbleness that lets a two-person shop compete with bigger firms. The filter for where to start, he adds, is usually just an interview with the owner about what growth would actually look like for them.

What the automation can't do

The clearest evidence for keeping a human in the loop, in Rock's account, is what happens when owners skip it. She's met business owners who ran their own SEO through a chatbot, did everything it suggested, and watched nothing improve. "They say, I did everything Claude told me to do, and it hasn't moved the dial, or it moved it in the opposite direction," says Rock. Often the damage traces to spinning up a pile of AI-generated pages with none of the business's own voice, which is exactly the build backfiring that pulls a site down rather than up.

What AI can't do yet, in her framing, is strategize for a specific business and its specific situation. The tools accelerate execution, but the call on what to pursue and what a metric really means is human work, and it's where a partner with a strong business mind earns their keep. Some of their clients are AI-averse and proud of not using it in their own operations, and RevRocket will still deploy it behind the scenes to get that business found when a prospect asks an assistant where to go, precisely because someone is steering it toward an outcome the owner actually wants.

Revenue over vanity metrics

The measure that governs the work is revenue, not traffic. Clicks and visits are pleasant numbers, but the company is named for the bottom line, and Levy is pointed that the spend has to justify itself against what the client earns. "If we're driving revenue for you, the investment you're making in our services becomes irrelevant," says Levy, who frames the job as looking at a client's business dynamics and being honest, month to month, about what's tied to revenue and what isn't.

That lens is also how he explains why the work can't be switched on and off. Paid ads stop producing the moment the budget does, while organic search compounds over time. "When you turn the ads off, the traffic is just gone, and it's not compounding anymore," says Levy. RevRocket doesn't hold clients to contracts, on the logic that results, not lock-in, should keep a business paying, and that a client seeing real return tends to accelerate rather than cancel.

The wrong-customer problem

Volume without aim is the failure Rock sees most, because more content and more ads only pull in more of the wrong people. She points to a client in Arizona selling extremely high-end mattresses, some ranging well into five figures, who kept getting sent shoppers hunting for a mass-market brand. "You can make all this content and run all these ads, but all you'll do is flood the zone with the wrong customer," says Rock. The traffic looked like a win and did nothing for the business.

The craft is in targeting precisely, which is why she pushes back on the instinct to spin up hundreds of articles to claim keywords. Producing content at volume is trivial now and beside the point if it draws random visitors instead of buyers. "We create content and make sure we're targeting exactly the right customer, not random people," says Rock, describing work like taking a local museum client toward a national audience by aiming the reach rather than widening it.

Taking on the incumbents

Where the agentic tooling gives a small shop real leverage is in exposing what larger agencies have missed. RevRocket's platform surfaces the technical weak spots in an incumbent's work, which becomes the trust-builder with a prospect. Rock describes prying loose a long-standing web-development agency after uncovering near-fatal errors the incumbent didn't know it had made. "Our platform uncovered this, and they were like, how did you figure this out?" says Rock. Sharing just enough of those findings, she says, is often what wins the room.

The result is a two-person operation delivering what a much larger entity does, and in one case a prospect who had a separate web company and PR firm came back asking RevRocket to fold all of it into a single proposal. What keeps that from feeling like pure automation is the deliberate human element, since Rock notices real AI fatigue among the people they pitch. "There's always going to be a thirst for the human touch," says Rock. "Not everything can be AI." The system, in their telling, is what runs agentically; the accountability for whether it's working stays with them.