Why AI companies are hiring for the jobs they’re supposed to replace

The problem is not that AI companies are hiring marketers. The problem is what that hiring pattern reveals about where value still sits. Anthropic has posted an SEO Lead role, Meta has shown an SEO Strategy Lead role, and OpenAI has an opening for Growth – SEO, CRO and Web Strategy. Those are not back-office support jobs. They are positions that sit close to distribution, narrative, and conversion. If the companies building the tools are still paying humans to own those decisions, that should tell the rest of the market something uncomfortable: the thing being automated is not the same thing as the thing that creates advantage.

Most teams misread this as hypocrisy. They assume the companies making AI tools should be the first to prove that content strategy, SEO, and web growth can be fully replaced by software. But that is a shallow reading of how operating systems work inside real businesses. Tool makers are not trying to eliminate judgment. They are trying to compress execution. That difference matters, because once execution gets cheap, the scarce resource stops being output and starts being decision quality. A model can draft a page. It cannot decide whether the page belongs in the market, whether it should exist at all, or whether it fits the long-term shape of the business.

That is why the hiring signal matters more than the rhetoric. If an AI company believed “just generate more content” was a winning strategy, it would not need a senior SEO lead or a growth strategist. It would simply let the models produce pages, test them, and scale volume. Instead, these companies are paying for people who can connect search, site structure, experimentation, product messaging, and measurement into one operating system. They are buying judgment because judgment is what determines whether the cheaper execution layer actually creates durable advantage. A lot of companies are learning this in reverse, after they cut the people who were making the important calls and kept the tools that only made those calls faster.

The non-obvious claim here is that AI does not make marketing less strategic. It makes the strategy more visible. Once anyone can produce a draft in a minute, the real distinction becomes whether the draft belongs to a coherent demand story. That is why the highest-value work shifts toward choosing what to write, what not to write, how to sequence the work, and how to make each piece compound into the next. The content itself is no longer the moat. The moat is the operating logic behind the content: the search map, the narrative, the internal linking plan, the conversion path, and the discipline to keep those pieces aligned over time. If that sequence breaks, you usually see the symptoms first as weak progression quality rather than weak traffic quality, which is why this pairs directly with B2B lead quality problems usually come from funnel leaks, not traffic.

A useful diagnostic is to ask three questions in plain language. First, does this work decide direction, or does it simply produce output? Second, does it compound across many future decisions, or does it disappear once shipped? Third, does being wrong here create a small formatting problem, or does it distort how the market sees the business? If the answer is direction, compounding, and high-cost error, then the work still needs an owner who can make trade-offs with context. AI can assist that owner, but it does not replace the ownership itself. This is the part many teams miss when they evaluate AI through the lens of labor reduction alone.

That diagnostic also explains why some companies are shrinking content teams while others are adding strategic roles. The firms cutting headcount are usually treating content as a production line. They believe the job is to fill a calendar, publish more assets, and let traffic sort itself out. The firms hiring SEO and web growth leaders are treating content as a system of bets. They want someone to decide which pages deserve attention, which topics align with product demand, which experiments should run first, and what metrics will prove that the work is moving the business forward. The second model is slower to set up, but it compounds. The first model is easier to start and harder to recover from.

You can see the difference in how these companies describe the roles. OpenAI is not asking for a generalist who can “use AI to write faster.” Its growth opening is about SEO, CRO, site architecture, experimentation, and web strategy. Anthropic’s SEO Lead role is tied to technical SEO infrastructure across claude.ai, docs.anthropic.com, and anthropic.com. Meta’s SEO Strategy Lead role, as mirrored publicly, is framed around search visibility, traffic, and conversions across B2B and B2C properties. In other words, the companies are not hiring for content volume. They are hiring for coordination across systems that influence demand. That is a very different job.

This is where the practical read becomes useful for any marketing team. If your current process starts with “what can we publish this week,” you are already optimizing the wrong layer. AI will make that mistake cheaper, not better. If your process starts with “what should the market understand about us, where is the evidence that it is not understood yet, and what sequence of pages or posts will move that belief,” then AI becomes a multiplier. It can accelerate drafts, summarize source material, suggest internal links, and surface gaps. But the human work stays on the more expensive side of the problem: deciding what matters and why. For teams building the operating model behind that work, this usually sits alongside Measurement and Attribution Consulting and a clear channel strategy sequence.

There is a counterargument worth taking seriously. It is possible that these roles are transitional, not permanent. Maybe AI companies are hiring SEO leaders because search still matters today, not because it will matter in the same way five years from now. That is plausible. Search behavior is changing. AI answers, answer engines, and new distribution surfaces will continue to shift where attention goes. The trade-off is that even if the channel changes, the underlying work does not disappear. Someone still has to own narrative, information architecture, experimentation, and measurement. The channel may move. The need for judgment does not. What may shrink is the number of people needed to execute the work, not the need to decide what good looks like.

The real trade-off is not “humans or AI.” It is fixed cost versus compounding quality. If you keep strategic people in the loop, you accept a higher steady cost in exchange for better sequencing, cleaner positioning, and stronger feedback loops. If you remove them and rely on AI alone, you lower the apparent cost of production, but you also increase the odds that the organization publishes more and learns less. That trade-off is easy to miss in the first quarter because the charts often look fine at the page level. By the time the deeper weaknesses appear, you have already built a large body of content that does not quite say anything specific enough to matter.

The practical implication for this quarter is simple. Do not evaluate your content operation by output alone. Evaluate it by whether the work is creating a clearer market position and a better conversion path. If you have not already done so, assign one person to own the search map, the narrative map, and the experiment calendar. Let AI accelerate the tedious parts: first drafts, rewrites, summaries, internal link suggestions, and content QA. Keep the human accountable for topic selection, prioritization, and the interpretation of results. If the strategist cannot explain why a piece exists and what business outcome it supports, the piece is probably not strategic enough to ship.

That same logic should shape hiring. The next version of a marketing team is not a room full of people manually producing more pages. It is a smaller group of operators who can define the business problem, use AI to reduce execution friction, and make decisions fast enough to keep the system learning. The companies that understand this will look less impressed by raw output and more interested in whether a page changes behavior. They will care about whether a post brings in the right traffic, whether a page helps a visitor self-select, and whether a conversion path makes the business easier to trust. AI helps with all of that, but only after someone decides what the system is for.

That is why the contradiction is not really a contradiction. It is a signal. The tool makers already know that execution gets cheaper before judgment does. They are hiring accordingly. If you work in marketing, content, or growth, the question is not whether AI will take over more tasks. It will. The question is whether your job currently sits in the part of the stack that gets cheaper, or in the part that gets more valuable when the cheap work is removed. If it is the second one, your task is to make your judgment visible, repeatable, and hard to ignore.

Next step

If you want a clear read on where strategy ends and execution begins in your own content operation, send me the last three roles you would hire for marketing or growth. I will tell you which ones should stay human, which ones can be partially automated, and where you are probably paying for volume when you should be paying for judgment.

Evidence and references

OpenAI Careers: Growth – SEO, CRO and Web Strategy shows a public role centered on SEO, CRO, site architecture, experimentation, and web growth. It is a direct example of an AI company treating strategy as a separate layer from execution.

Anthropic SEO Lead on Weekday mirrors a public posting describing ownership of SEO strategy and technical SEO infrastructure across Anthropic web properties. The job is framed as infrastructure and judgment, not generic content production.

Meta SEO Strategy Lead on Ladders mirrors a public listing focused on search visibility, traffic, conversions, and team leadership across Meta web properties. It reinforces the same pattern: AI-adjacent companies still hire humans to own the system around the model.

OpenAI Careers search page is also useful context for how the company currently structures roles across growth, marketing, product, and operations.