SEO Revenue Drivers: What Still Works to Drive Pipeline and Revenue

Let’s start with something I keep running into when I open client dashboards this year: organic sessions are down and organic-influenced pipeline is holding steady or growing. The two lines used to move together and they stopped doing that somewhere around the middle of last year, which makes for a confusing conversation when the quarterly review comes around and somebody asks whether the channel is still working.

Ahrefs published numbers from their own analytics in June 2025 that show the same thing from the inside, and they’re the cleanest illustration I’ve found: AI search was 0.5% of their traffic and 12.1% of their signups. Half a percent of the visitors producing an eighth of the revenue events. 

Their sample is one company in a category with unusually high AI adoption, so I wouldn’t treat that ratio as a benchmark for anybody else, but the direction has held up in every study published since.

I want to spend this piece on what that shift means for where the money goes, because there’s a version of this conversation that ends with “cut the content budget, traffic is dying” and I think that reading gets the mechanics backwards.

What’s actually happening to the buyer?

The short version is that somebody else is doing the early work now.

It used to be that a buyer with a problem would search, land on four or five sites, read around, and slowly assemble a shortlist on their own. That whole process happened in public, on your site, and every step of it showed up in your analytics. 

Now a good share of buyers ask an LLM instead. The AI works out which category solves the problem, decides which companies belong in the answer, explains why, and hands the buyer a shortlist. 

The buyer clicks through to verify the recommendation, which is why fewer people arrive and the ones who do arrive so much further along.

The framing I’ve found most useful for this, and the one that tends to land immediately with sales leaders, is that the assistant is doing SDR work. It’s qualifying, it’s educating, it’s building the shortlist, and it’s warming the buyer up before the handoff. That’s the job description.

And here’s what follows from it. Every sales leader knows that a rep is only as good as the material you give them. A rep who understands what the product actually does, who has honest answers about where it fits, and who can speak to how it stacks up against the alternatives will represent you well. 

But a rep with nothing to work from will improvise, hedge on the specifics, and often end up describing the competitor they happen to understand better. Nobody blames the rep for that. You fix it by fixing the enablement.

LLMs work exactly the same way, with two differences that matter: 

1. It has no loyalty to you
2. And it can’t message your product marketing lead when something is unclear. 

It builds its recommendation out of whatever it can find about you, and if what it finds is thin, it fills the gap with a two-year-old review or a competitor’s comparison page written specifically to make you look worse. So the enablement problem is the same problem, just with a different solution. 

So what does that mean for your site?

It means product, industry and feature content became a revenue driver again, which is a strange sentence to write because it sounds like something from 2015.

• What your product does, in specific terms
• What it doesn’t do, said plainly, because a recommendation that oversells you produces a demo call that goes nowhere
• Feature detail precise enough that somebody could actually tell you apart from the alternative
• Pricing on the page rather than behind a form
• Integrations named individually instead of gestured at with a logo wall
• Prerequisites and limitations written down somewhere other than a sales rep’s head

Comparison pages sit here too, and the ones that work are the ones that survive scrutiny, including the dimensions where a competitor legitimately wins. 

That feels counterintuitive until you watch a buyer read one. Honest comparisons get trusted, and they get picked up as source material far more readily than a page that only makes the case for you, because a system assembling a recommendation is looking for something that reads like an assessment.

Documentation deserves a specific mention because it’s the piece I see left out of these conversations most often. Official docs and help centres gained substantial ground in the recent reshuffle of what ChatGPT treats as a citable source. 

If your documentation sits under support, has no owner in marketing, and hasn’t been revised since 2023, that’s a genuine win sitting there waiting for you. 

The nice thing about this whole layer is that one asset serves two audiences. The assistant reads it to build the recommendation, and the human who clicks through reads it to confirm the recommendation was right. You’re enabling the rep and equipping the buyer with the same page.

Where does original data fit?

This is the investment I’d protect hardest, and the one that gets cut first because it looks like a research expense rather than a pipeline expense.

Rankings get reshuffled. Citations get reweighted, sometimes dramatically and with no notice. The one position that survives all of that is being the only place where a piece of information exists. 

If you publish the benchmark for your category, or a survey nobody else has the audience to run, or an analysis built on data that only your product generates, you become a source rather than a participant. 

Everyone writing about the topic has to reference you, and that reference accumulates into the kind of coverage these systems read when they work out which companies belong in an answer.

One good dataset a year, promoted properly, keeps earning citations across every platform simultaneously, because there is no substitute source available to swap you out for. It’s the closest thing to a durable moat that content marketing offers right now.

Does broad content still earn its place?

Yes, and I want to be clear about it because the conversion data on AI referrals has been used to argue the opposite.

Those conversion multiples measure the last mile. They describe what happens when somebody arrives already convinced, and they tell you nothing about what did the convincing. 

The 95:5 rule from Les Binet and Peter Field is the frame I keep coming back to here: at any given moment roughly 5% of your addressable market is actively in-market, and the other 95% will get there later, on a schedule set by something happening inside their business rather than by anything you published. 

Your product pages compete for that 5%. 

Everything you publish about the problem itself is there so that when somebody from the other 95% finally has the problem, your name is already sitting inside the category in their head.

There’s a second reason that’s become more mechanical. Google and the LLMs assemble a profile of your company from everything they can find, and that profile is what decides whether you appear when someone asks which company solves a given problem. 

A company publishing genuinely useful material about the problems it works on reads as an authority on those problems. A company whose entire site is about its own product gives the systems very little to place it with, however good the product pages are.

The thing I’d change is how this layer gets judged. Measure it on citations earned and on branded search movement, because measuring broad content on last-touch conversions will make perfectly good work look like a failure and lead you to cut the thing that feeds everything else.

And PR, mentions, community?

Same profile, built mostly from what other people publish about you. Press coverage lands in it, review sites land in it, and so do podcast appearances and the conversations happening in the communities where your buyers already spend their time.

What I’d emphasise here is that distributed presence is what makes the profile resilient. When ChatGPT’s source mix shifted in August and Reddit citations collapsed in six days, the companies that had built genuine presence across a range of places barely felt it, because the profile absorbs the loss of any single input. 

Coverage concentrated in one place is a tactic with an expiry date somebody else controls.

Why does none of this show up in the organic report?

Because most of what organic produces gets collected by other channels, and this is the part I’d make sure lands in a budget conversation.

Someone reads an article of yours, doesn’t buy, and quietly enters your retargeting pool. Three weeks later they convert from a retargeting ad, and paid takes the credit for an audience that organic assembled. 

Example 2. Someone else meets you in a community thread, searches your brand name rather than clicking anything at the time, and lands on a branded campaign that posts a beautiful cost per acquisition precisely because the demand arrived pre-formed. 

And I have another one. Your email list fills up with people who found a blog post first. The account-level signals your sales team scores against light up because somebody spent an afternoon reading your comparison pages. None of that appears in an organic report, and all of it is organic doing the work.

The practical fix is a reporting one:

• Report organic-influenced pipeline alongside organic-sourced, with both definitions agreed with RevOps. 
• Put citation share next to rankings so people can see how often you appear inside the answers buyers actually read. 
• And if you want one piece of hard evidence, see how paid branded campaigns are behaving and how much of the total SEM results come from them. 

The takeaway

So, what still drives pipeline? Product and feature content that gives the systems recommending you something real to work with. Data that makes you a source instead of a participant. Broad content that puts you in the category before the buying window opens. Press and community presence that keeps the profile strong across whatever platform happens to be routing attention this quarter. 

Every one of those was already good practice before AI search existed, which is either boring or reassuring depending on how your last quarter went.

What did change is where the value lands. It arrives as a warmer buyer, a cheaper branded click, a retargeting pool that converts, a sales call that opens with the prospect already knowing what you do. 

If your reporting only counts sessions and last-touch conversions, all of that reads as decline, and the decision that follows from that reading is to cut the thing producing it.

So before the next budget conversation, go find out where your buyers actually came from. Put “how did you first hear about us?” into every discovery call and every onboarding form, then compare a month of open answers against what your attribution model claims. 

In my experience the gap is ENORMOUS, and it’s the cheapest research a marketing team can run. You’ll walk into that meeting with your own data instead of an argument about how the funnel has changed, which is a much better position to be in.