Is Ranking on Google Still Worth It?

Ranking doesn't matter anymore, it's all AI now, stop wasting money on SEO. That gets said with total confidence, and it's wrong in a specific way worth correcting. I think that's wrong, and wrong in a specific, correctable way rather than a vague "well it's complicated" way. Ranking did not stop mattering. What stopped being true is that ranking is the finish line.

Start with the part that hasn't changed. Before an AI answer engine can ever cite you, it has to already have you sitting in an index it can search. If your page isn't crawled, isn't indexed, or is buried so deep that classic search can't surface it when the system goes looking, you were never a candidate for citation in the first place. That part of the game is exactly as real as it was five years ago. Being indexed and reasonably rankable is the entry ticket, not a nice-to-have. An invisible page cannot be cited, no matter how well you write it. Ranking is the gate you clear just to be in the room.

Here's the part that actually changed, and it's the part that trips people up. Being in the room isn't the same as being the one who gets to answer. Once your page clears that gate and lands in the pool of things the machine could pull from, whether it actually gets cited from there is decided by something else, and it happens well downstream of where you sit on the results page.

Google has said outright that its AI Mode doesn't answer your one question, it breaks it apart and issues what the company itself calls "a multitude of queries simultaneously," with its deeper research mode running hundreds of them behind a single prompt. So the system isn't asking "who ranks well for this topic." It's asking dozens of narrower questions and looking for whatever answers each one best. Retrieval at that stage doesn't happen at the level of your website, or even a whole page. It happens at the level of a passage, described by engineers building these systems as operating on the content at, in their words, passage or sub-document granularity. The chunk, not the domain, is the unit competing.

Then comes attribution, and this is the part almost nobody outside the companies building these systems talks about. Google holds a patent on generating these answers, US11769017B1. It describes a system that writes a claim first, then goes looking for a passage close enough in meaning to back it up, and links that one. In plain terms, the machine writes the sentence first, then goes looking for a passage close enough in meaning to back it up, and cites that one. Which source gets named is decided by how closely a passage matches a specific claim, not by where the page containing that passage ranks.

There's a real number behind how sharply this can diverge from rank, and I want to be precise about where it comes from before I quote it. In a large controlled measurement of what actually gets cited versus what ranks, pages ranking fifth on a query picked up citation share while the pages ranking first on the same query lost it. Keyword-stuffed pages scored below the baseline entirely. Same query, same results page, wildly different outcome once you look at which passage actually answered the specific sub-question being asked. I'd call the direction here solid: it comes from a large measurement wave and lines up with everything else I've seen about how these systems select sources. I would not treat the exact percentages as something you should expect to reproduce on your own query tomorrow. Treat it as proof the gap between rank and citation is real and can be large, not as a formula.

What this means in practice is that the question worth asking on a Monday isn't "am I ranked." It's "for the exact question my customer is asking, do I have a passage that answers it plainly?" Those are different questions with different answers. You can rank well for your category overall and still lose every sub-question that drives a purchase. Nobody wrote a clean, self-contained passage answering "how much does delivery cost" or "what happens if I miss a payment."

None of this is an argument to stop doing the boring ranking work: technical health, real content, a site that loads and gets crawled, earned links that reflect real relationships. Skip that work and you're not in the index at all, and none of the rest of this applies to you. But treating page one as the finish line, and stopping there, misses the actual game now being played downstream of it. A smaller business with a thinner backlink profile can win the specific sub-question a much bigger, better-ranked competitor never bothered to answer clearly, because relevance to the exact question is what gets rewarded at that stage, not accumulated domain authority.

If rankings and citations were the same mechanism, they could not move in opposite directions on the same page in the same quarter. They can. I don't think that contradicts anything I've said here. I think it confirms it. If ranking and citation were the same mechanism, they couldn't move in opposite directions on the same page. The fact that they can, and regularly do, is the clearest evidence I have that you're now managing two separate assets, not one. A rank tells you the page is healthy and findable. It tells you nothing about whether the specific sentence a customer needs is sitting inside it, stated plainly enough to be lifted out and quoted.

So the actual audit worth running isn't "check my rankings" and it isn't "check my AI visibility" as some separate, disconnected exercise. It's both, together, on the same page. Confirm you're still indexed and reasonably visible for your category, because that's the gate. Then, separately, go read your own page the way a machine would, looking for the exact sentence that answers the exact question your customer is asking, and ask whether it's there at all, and whether it's buried three paragraphs into something else. Almost nobody has done the second half of that audit, because for fifteen years the first half was the whole job.

Ranking gets you retrieved. It doesn't get you named.