I Used To Think Ranking Was The Point

Matt Emgi / October 26, 2026

Act 1: What I believed

I got into this industry as a content writer.

I was writing my own blogs, learning about SEO one post at a time, trying to work out why a couple of them ranked and most of them didn’t. I think about this: it was only five years ago. Wow, so much has changed since then.

Back then keyword research meant opening a tool. Ahrefs, Semrush, Keywords Everywhere, then Answer The Public, and people will remember this, you wrote content by looking at the People Also Ask questions. The model was simple: answer every question a buyer could have, in more depth than the competitor. It genuinely worked for a long time.

And authority meant one thing. Backlinks in relevant publications. Industry press. PR. You acquired them, your numbers went up, your pages moved.

So ranking was the scoreboard, and I treated it as the point.

Then in December 2022 Google added a second E to E-A-T. Elizabeth Tucker announced it on their own blog, and the wording was blunt: “Now to better assess our results, E-A-T is gaining an E: experience.”

The question they attached to it is the one I’d pin above my desk now:

“Does content also demonstrate that it was produced with some degree of experience, such as with actual use of a product, having actually visited a place or communicating what a person experienced?”

A big part of E-E-A-T that a lot of people don’t admit is that in practice, a huge amount of it comes down to backlinks and the authority of the site. It always has. How else was Google supposed to tell quality apart at scale? It’s why parasite SEO can even exist: if links didn’t matter, we’d all be competing on content quality alone. We’re not. That’s the uncomfortable bit. But the new E was the one that actually mattered, and most of us skimmed past it to get back to link building.

First-hand. That’s the bit.

Act 2: What broke

I remember creating content years ago with RankIQ, Brandon Gaille’s tool. He did a podcast about blogging, and I thought he was a cool guy, and he had built this thing where you could see all the LSI keywords and then include them in your page.

The process was honestly agonising. Write the thing, paste it into the tool, watch the score, go back and sprinkle in whatever the machine said was missing, like seasoning a soup you were never going to eat.

And yes, the content ranked.

But content can rank and it doesn’t mean anything.

Ranking was never the point. It was a proxy we all agreed to treat as the point, because it was the thing we could measure. Traffic is the same kind of proxy: objective, clean, a number of visitors. The messy part is whether any of it turns into business, and that depends on the offer, the sales process, the conversion work. So we reported the clean number and let everybody assume the messy part followed.

That’s the whole argument of this piece, really: this industry keeps confusing the metrics we can measure with the outcomes businesses want. Everything below is that one mistake, repeating.

I lived this one personally. Back when I was blogging I had a page doing 10,000 visitors a month about how to host a foreign exchange student. Okay, maybe I’m exaggerating. I think it was 5,000. The site was not even a foreign exchange site. Chasing that traffic felt like the smart move at the time, and it made no difference: barely any enquiries, and the few we did get were from people who genuinely wanted to host foreign exchange students, which was not a thing we did.

What I didn’t understand then is what information gain actually means. Not longer. Not more comprehensive. Unfakeable. Real world expertise and real world examples that could not have been invented by someone who hadn’t been there.

Google holds a patent on something close to this, and it’s worth being precise, because the industry has been loose with it for years. The honest version: it’s written for automated assistants, re-ranking references after somebody has already read something. Google has never confirmed it runs in organic ranking, and anybody selling you an “information gain score” is selling their own metric, not Google’s.

But it tells you what Google’s engineers consider valuable: new information, not more information. That distinction is the whole game.

So that was the first time I watched this industry mistake a proxy for the point.

Here is the second one, and it’s happening right now.

A lot of people currently believe the whole point of AI search is to get cited. There’s a serious amount of money being raised by companies that will track your mentions in ChatGPT and hand you a number.

I said this in a couple of the marketing communities I’m in: surely we can agree that prompt tracking and AI visibility is not how we measure the success of this work. It’s the same as traffic. It’s a vanity metric.

I’ve been living a version of this recently with a security client in the US. Objectively, the work was a link building campaign: improve the rankings of specific pages. The rankings were fine. Their visibility scores in HubSpot’s AEO tool were actually improving. And fundamentally they weren’t getting enough leads from search, which is the thing they actually care about. However good the backlinks are, the backlinks aren’t going to pay the bills.

I watch how hollow a citation can be in my own audits every week.

A few weeks ago I ran a live test for a cash flow software company. Good product, and credit where it’s due: a well built site with proper, detailed comparison pages. That is exactly why they showed up. I asked ChatGPT for the best cash flow forecasting tool for a small business, and there they were, named, with a nice little verdict attached. Lovely. Looks like a win.

Then I read the citations. Nearly every source in the answer was their own website. Their comparison pages, their own listicle. And the same answer recommended one of their competitors, with their listicle cited as the source for it.

Their content was good enough to build the whole answer from. It still didn’t make them the recommendation.

And it won’t be the same answer next week. Run these tests enough times and you learn that who gets cited moves around constantly. What holds steadier is who gets recommended, and that’s anchored somewhere else entirely:

• The lists you’re in.
• The competitors’ pages that name you.
• How much of the web mentions your brand at all.

Lily Ray found the mechanic at scale across 100 B2B software queries, and her September re-run found AI Overviews excluding the self-promoting brand 83% of the time.

Those are two completely different things. The only one that matters is being recommended.

We’ve run this kind of live test for over 150 companies now. Across 138 we analysed properly, at least 40% were provably absent by name from their own category’s AI answers. In our latest batch of 25: 12 named, 13 absent, and several of the named only because a rival’s listicle happened to include them.

You can be excellent at the old scoreboard and invisible on the new one, and the two new scoreboards don’t even agree with each other. One number cannot describe that.

I’ve sat with these dashboards with clients and sales prospects, Peec and Profound mostly, and the pattern repeats: they’re tracking a handful of bottom of the funnel prompts and comparison prompts, which of course cite the brand, because the brand name is in the prompt. Charging a retainer to confirm that ChatGPT mentions you when somebody types your name in is a hell of a business model.

So when a dashboard hands you one AI visibility number, ask:

• Which engine is it describing?
• Which intent?
• Averaged across what?
• And before all of those: who decided the prompt set?

Somebody chose those prompts, and the entire number rests on the claim that they’re what your buyers actually ask. Nobody can verify that.

That’s the same trap as the conversion stat everyone is quoting right now. You’ll have seen that AI traffic converts around four and a half times better than organic. Marie Haynes’s data says exactly that: 13.8% against 3.02%, across the 55 of her 100 sites that tracked conversions. It’s real. I see it with my own clients too, across a much smaller sample, and even on my own site: tiny amounts of AI traffic, converting remarkably well, generating actual leads.

But The SEO Works benchmarked 95 sites and segmented it. For B2B products, AI traffic converted at 0.62x. Worse than organic.

The headline multiplier is a B2C-weighted average that inverts for the exact segment most of the people quoting it are selling to. The number stops being true the moment you stop segmenting it. We keep doing this.

The decoupling evidence is contested, and it would be dishonest not to say so. Scrunch found the effect much weaker across 10,000 URLs. And on the wider measurement fight, Mike King’s rebuttal to the AI traffic studies argues “every one of these studies grades a brand channel with a traffic ruler.” He’s right about the ruler, and the metric is still wrong: he’s defending the channel, I’m saying the number doesn’t measure the thing that makes anybody money. What even is it? A share of voice? Visibility of what, to whom, measured how? We’re invoicing against a metric nobody can define.

Act 3: What I think now

So what does organic growth really look like for a brand now?

I’ve started calling it organic positioning, and honestly, the name is not the thing I care about. Positioning rather than brand, because brand gets people thinking about logos, whereas positioning dials people into their customers, what they sell and who they sell to. It’s semantics, really.

What I care about is the work underneath it: organic growth grounded in a framework. Content with genuine information gain. Authority earned by being expert-led, by making excellent content that isn’t easily replicated. And SEO as the structure: how people actually search, optimised for deliberately. We don’t start with SEO data. We start with how we want to sell the brand, the problems, the ICP, the use cases, the industries. Then the SEO data tells us the search intent, and we build the pages.

Here’s the distinction I’d sharpen, because I blurred it myself for years. Rankings and citations measure discoverability: can you be found. Recommendations measure consideration: do you make the shortlist. And qualified pipeline measures whether any of that visibility actually shaped a buying decision.

Even recommendation isn’t the finish line. If a brand gets recommended to twenty people a day and it never shapes a buying decision, what was the point? It’s the virality question from social media. Do you want views, or do you want money? Businesses want money.

But you climb the ladder in order. Discoverable, then considered, then chosen. Most of this industry is still optimising rung one and invoicing for it.

Here’s the number that changed how I think about my own job. Ahrefs studied 75,000 brands and correlated a set of factors against appearing in Google AI Overviews.

Unlinked mentions of your brand correlate roughly three times more strongly than the backlinks I spent five years learning how to acquire.

It lines up with the audits. The companies that show up in AI answers are the ones the web already talks about. The ones that don’t usually have the backlinks, the content, and nobody saying their name anywhere a machine reads. Links are real signals, but they’re proxies. The thing itself is whether you are becoming a better known brand in your specific industry.

And I’ll flag a tension here rather than hide it: our own citation gap study found raw web mention counts barely predicted ChatGPT citations, while Ahrefs found branded mentions strongly correlated with AI Overview presence. Different engines, different measures, different answers. Which is, again, the argument: no single number describes this, and anyone selling you one is selling the measurable over the meaningful.

Some people are calling this fame engineering. Andrew Holland has written about it as the key to generative engine optimisation, and to his credit he’s honest that the underlying idea isn’t his: he points to Paul Feldwick, and the serious evidence base is Les Binet and Peter Field’s IPA work across 996 campaigns, where fame came out as the single most efficient effect in advertising.

Think about how you actually buy. The last thing I paid for was an Instagram growth service for my own personal brand. I found them through an AI recommendation, as it happens. What convinced me was the way they spoke to me and the copywriting on their website, then a sales process and results that backed it up. The recommendation opened the door. The brand walked me through it. The biggest thing is the presence of the brand, and very often the presence of the founder.

I have to trust you. Attention got cheap and trust got expensive. That’s the whole shift.

It’s why founder brand has become so big, and personal brand generally. It’s why I’m sat here writing this. We want to share our way of thinking, and the people who do that get bought from.

Which is also why every page has a job: to be a touchpoint. An email signup, a lead, or a seat in a consideration set that might not surface for months.

This week I had an inbound sales call come in from a funded company. They’d read one really old case study of ours, asked ChatGPT for a company like us, and booked the call. I’d never have predicted that page would do it, and no prompt tracking dashboard would have told me to build it. Kevin Indig ran 632 tests looking for a reliable link between AI visibility metrics and later branded search or direct traffic, and found nothing detectable without self-reported attribution. Both things are true at once: the influence is real, and the dashboards can’t see it.

One of our clients, a recruitment business, now gets 26% of its inbound leads from people who say an AI recommended them. A quarter of their lead flow, arriving through a door no analytics package watches. It is not about getting found in AI. It’s about being found in the searches that are truly relevant, by the people with the exact problem you solve.

SEOs have always had an attribution problem. Marketing was never about last touch; it was always about influencing a decision across touchpoints, and we knew it, even while billing as though we didn’t. A blog doesn’t convert like a landing page. Different job. It’s all one thing.

On measurement, Aleyda Solís has the most honest framework I’ve seen. Three layers:

After Aleyda Solís’ measurement framework.

Her words: “Not perfect attribution, it’s an honest reporting model.” That’s the right posture. You can’t claim credit you can’t prove. You also can’t dismiss influence you can’t measure. Hold both.

And there’s one more thing worth watching that almost nobody mentions: accuracy of representation. I have a client who gets recommended constantly, which sounds great, except they’re always recommended as the best low-cost option, and that bugs them, because they don’t want to be known as the low-cost option. Fixing that is positioning work: the consistency of how third-party listicles describe you, how you talk about yourselves on your own website, what the engines keep picking up. Being recommended wrongly is its own kind of invisible.

Act 4: So this is how I’d actually do the work

Let’s make it concrete. Say a cold calling software company came to me.

The first part of the work is not SEO at all. It’s positioning and messaging. We use SEO as a way to structure the argument.

Here’s how that starts for us. A questionnaire before anything else: who converts, what they pay, which deals you lose and to whom. Then time with the founder, because the founder is usually where the experience lives. The second E in E-E-A-T is sitting in somebody’s head: the war stories, the opinions, the things nobody has written down. Our job is extraction. We interview, we record, we mine the sales calls and support tickets and podcasts, and that becomes the raw material every page is built from. Then SEO becomes the engine that gives it structure.

A real example. One of our clients is a LinkedIn automation tool. The keyword data told us which gaps existed, fine, that’s table stakes. It was the founder’s backstory, in our interviews, that gave us a genuinely unique angle for a service page they had never built. The data said the keyword made sense. The founder’s experience made the page something a competitor can’t copy. And beyond ranking, it’s one more service an LLM can pull from when it’s working out what this company is. I believe in that.

So we start with the central storyline. Who is this brand, what does it replace, who is it for, why is it better. You do that exercise properly and you write it down, because every page you build afterwards inherits it.

Then the second question: what structure is optimal for capturing the most relevant people who have these problems?

That starts with the ICP and with voice of customer. Sales call transcripts. Support tickets. Industry podcasts. People complaining in forums.

I recommend this to everybody and almost nobody does it. You take the actual words buyers use and you structure the work around them.

For cold calling software that looks like:

• Features somebody actually wants. Power dialling, call recording, CRM sync. Not “our platform”.
• Use cases by industry. Cold calling for recruitment is a different job to cold calling for insurance.
• The problems that drive the purchase. Not enough leads. Connect rates falling. Reps quitting.
• The comparisons people run when choosing. Alternatives pages, best-of lists, head to heads.

Each page gets one job and one intent. Stop building pages that try to do three.

Mechanically, the page map starts with the sitemap. Pull it, then audit every existing page properly:

• Are pages cannibalising each other?
• Can some be merged?
• Does the content actually match the H1?
• What’s missing?
• What is so unfinished it should just be deleted? Deleting content is not a bad thing.

That audit reveals the keyword gaps, including the naming ones, where you call something one way and your industry calls it something else. Then the founder stories and the pain points, from the interviews, from Reddit research, from YouTube comment sections, get tied to the keywords that make sense. That’s the page map.

Then the real discipline is prioritisation, because there will always be more ideas than capacity: start with the bottom of the funnel pages that move the needle, and with improving what already exists.

Then the part most people skip: map the other companies in your category. Who’s big, who your actual competitors are, where you’re genuinely better, and where they’re honestly better than you.

You’re building a web, and this is where entity thinking comes in. Being named alongside the companies in your category is probably part of how a machine works out what category you belong to at all. I’d be careful how strongly anyone states that, including me: no primary source proves LLMs assign category membership from co-mentions. It’s a well-reasoned hypothesis, most people state it as fact, and it isn’t, yet. The practical version is older than AI anyway: Bill Slawski was writing about co-occurrence back in 2012.

And it helps to stop treating competitors as enemies. In the cold calling example, every other cold calling product is just another entity that exists near you. You don’t need to take it personally.

On search volume, one caution. Zero volume in a tool doesn’t mean nobody searches it: Google says 15% of daily queries are new. But Ahrefs tested the stronger version and found low-volume estimates roughly accurate. The case for these pages is intent quality, not secret traffic.

Then, for the specific pages with the highest likelihood of conversion, that’s where we deliberately point authority. Links aimed at those pages specifically, not spread evenly across a blog.

That’s the philosophy. Positioning first. Structure second. Authority aimed at what actually matters third.

And the question I keep coming back to: how do we create information gain in everything we publish? The answer is a research process. It means forcing somebody to have an opinion, which sounds obvious and is rare. It’s why great writers matter: not because they type words in a nice order, but because they have a way of thinking a machine categorically cannot do.

And it means covering the whole funnel, not just the commercial pages: the best-of lists and alternatives pages, yes, but also the content people read long before anybody is shopping.

Act 5: What I think the real problem is

The problem with most agencies is that they need to sell you a number of links and a number of pieces of content. That’s the product, because it’s the thing that can be counted and invoiced.

I know this from the inside, because I’ve sold those packages. The counting was for me, not the client. It made my service legible on an invoice. It never described whether the market knew who my client was.

SEO has an attribution problem, yes. But the way this gets bought is also the problem. There’s no point in an audit you file and forget. Action, and being in motion, is what causes a company to get results from search.

You don’t need a package. You need a position, and the motion to earn it: the pages that state your argument, the authority aimed at them deliberately, and your name appearing in the places your category gets decided, whether the thing reading those places is a buyer or a machine. That is where the ROI of SEO actually lives now: not in the metrics we can measure, but in the outcomes businesses want. Qualified pipeline, from being the brand that gets chosen.

That’s organic positioning. It’s what I’d buy if I were you.