GEO might give thought leadership the feedback loop SEO never really did

Thought leadership has always had a measurement problem.

We can measure the things around it. Traffic. Downloads. Coverage. Backlinks. Engagement. Brand search. Executive visibility. Sales usage. Event conversations. Sometimes we can run brand studies or track whether a particular phrase starts appearing elsewhere.

All of that can be useful.

The harder question is what actually caused what.

An organisation commissions research. The research becomes a report. The report produces articles, executive commentary, PR, social posts, webinars and sales material. Journalists reference some of it. Other sites link to it. Customers encounter the argument. Search engines index the resulting material. Six months later, perhaps the organisation looks more visible or more authoritative around the subject.

Was that the research? The links? The PR? The domain? Repetition? Brand recognition? A page rewrite? A change in search demand? A competitor going quiet?

Usually the answer is some combination of those things.

That does not make thought leadership ineffective. It makes learning from thought leadership unusually difficult.

SEO never really solved that problem for thought leadership either.

The familiar argument has always been something like this:

publish strong, original work → earn attention and links → build authority → improve search performance

There is plenty of logic in that chain. But it is still a long chain, and long chains are very easy to narrate after the fact.

I am starting to wonder whether GEO gives us a smaller unit of learning.

Not perfect attribution. I do not think that exists here.

Something more useful: a shorter distance between question, intervention and observable signal.

Shrink the unit of analysis

The mistake may be trying to measure thought leadership only at campaign level.

A research programme can contain dozens of claims, questions, pages, videos, interviews and audience needs. Asking whether the whole campaign improved AI visibility creates almost the same problem as asking whether the whole campaign improved SEO.

There are too many moving parts. But generative answer environments can sometimes be inspected at a much smaller level, so start with one question: what answer currently appears? Which organisations and sources are visible? What evidence do those answers rely on? Where is your organisation absent? More importantly, what is missing from the answer itself?

That last question changes the job.

Suppose an organisation has almost no visibility for an important audience question. The obvious response is to create a page targeting that question.

But perhaps the existing answers already explain the facts perfectly well.

Another summary of the same information might add almost nothing.

What the answer may lack is practitioner experience. Or original data. Or a useful example. Or a customer perspective. Or a distinction nobody has explained clearly. Or evidence the organisation already holds but has never connected to the question.

Now the intervention becomes more specific:

question → current answer → diagnosed gap → deliberate intervention → publish → observe the question again

That is a much smaller experimental unit than “thought leadership campaign”.

And therefore potentially a much better learning unit.

This is what I am trying to test

I have been experimenting with this in current work around Making Tax Digital.

The starting point was not a content brief. It was an AI visibility signal.

For one useful question, the organisation I work for was effectively absent from the answer environment I was examining. The obvious response would have been: we need content answering this question.

But when I looked at the existing answers, the factual distinction was already widely available. Government guidance and other accounting sources could explain it perfectly well.

The problem did not appear to be a lack of basic information.

So I started asking a different question:

What could we contribute that would actually make the answer better?

One possibility is practitioner evidence.

What do clients misunderstand? Where does the actual work appear? What gets chased, checked or corrected? What does somebody dealing with the issue every day wish clients understood earlier? How would they explain it in thirty seconds?

That turns a GEO signal into something much closer to a reporting brief.

And because I am attending Accountex North, there is suddenly a practical way to investigate it.

Instead of going to an event with a vague instruction to capture content, I can take a specific evidence gap into the room. Talk to practitioners. Test the assumption. See whether the supposed gap is real. Come back with evidence that may strengthen an article, a commercial page, a short video, social material — or perhaps demonstrate that no new content is needed at all.

Then we can look at the original question again.

That is the experiment.

The interesting measurement comes before the visibility result

It would be easy to define success as: did AI visibility go up? Eventually, yes, I want to know that, but it is not the first useful measurement. The first question is whether the GEO analysis changed an editorial decision.

Did it lead to a better interview question? Did it stop us producing a generic article? Did it expose evidence we already had? Did it identify something genuinely worth researching? Did it give a video team a better question to put to a practitioner?

That already tells us whether GEO is contributing intelligence rather than simply another optimisation checklist.

The next question is whether the intervention improved the answer we could publish.

Perhaps the practitioner interview reveals a misconception nobody else is addressing. Perhaps the assumption turns out to be wrong. Perhaps the strongest output is a 45-second video rather than a 1,500-word article. Perhaps the evidence belongs on an existing evergreen page rather than inside a new campaign asset.

Only after those decisions do I become particularly interested in what happens to AI visibility.

The measurement sequence becomes:

Did the signal change the question? → Did the investigation change the evidence? → Did the evidence change the output? → Did the answer environment change afterwards?

That feels more useful to me than jumping directly from “we made good thought leadership” to “our visibility improved”.

This still does not prove causality

There is an obvious problem with this argument.

Generative answer environments are volatile. Sources appear and disappear. Different models produce different responses. Retrieval changes. Search products change. Citation behaviour changes. A competitor publishes something. An authoritative external source updates a page.

So if an organisation moves from 0% visibility to 20% after making an intervention, it would be ridiculous to immediately claim that the new article or video caused a 20-point improvement.

That is not what this model offers.

What it offers is a better observation loop.

We know the starting state. We make a deliberate intervention based on a specific hypothesis. We inspect the state afterwards. Then we repeat the process.

One result is anecdote. Ten similar experiments start to become a pattern. Dozens of interventions, recorded properly, could begin answering much more interesting questions.

Do practitioner quotes appear more readily than generic brand copy? Does original research travel more reliably than interpretation? Do short direct-answer videos become useful citation sources? Does improving an existing authoritative page outperform creating a new one? Does customer evidence affect a different class of prompt from product information? Which interventions appear to do nothing?

The point is not to manufacture certainty from unstable data.

It is to accumulate organisational learning from repeated interventions.

That is also why GEO without a playbook matters to me. In a discipline this young, the useful capability may not be memorising somebody else’s rules. It may be building a disciplined system for discovering which rules survive contact with the work.

Thought leadership has usually measured assets. This could measure knowledge movement.

This may be the part I find most interesting.

Thought-leadership measurement often starts with the asset. How many people read the report? How many journalists covered it? How many executives posted about it? How many leads did it influence?

Those are legitimate questions, but they keep the unit of analysis centred on content.

GEO creates the possibility of looking at something slightly different:

Did a piece of organisational knowledge enter the wider answer environment?

That knowledge might begin as a statistic, a practitioner observation, a customer example, a useful distinction, a direct explanation or a proprietary argument.

If it begins appearing elsewhere — in AI answers, citations, external coverage, customer conversations or the language other people use — then we are observing the movement of an idea rather than merely the performance of an asset.

That feels much closer to what thought leadership is supposed to achieve.

It is also why I am increasingly interested in the evidence underneath content. The article is not necessarily the valuable unit. The valuable unit may be the claim or piece of evidence that can travel between an article, a video, a web page, an executive conversation and an AI answer while remaining recognisably useful.

That is the same reason the Evidence Engine starts before publication and the Content Activation Framework treats channels as different expressions of the same underlying knowledge rather than separate production jobs.

This changes the relationship between GEO and thought leadership

The common argument currently runs in one direction:

stronger thought leadership → greater authority → better GEO performance

That may well be true.

But as a working method, I find the reverse direction more interesting:

GEO signal → better question → better evidence → stronger thought leadership → measure again

Now GEO is not simply downstream of content.

It starts informing the research. It may change who gets interviewed. It may change what is asked at an event. It may reveal that a customer story contains evidence which should be connected to another subject. It may tell a video team which audience question deserves a direct expert answer. It may reveal that a commercial page needs one concise explanation rather than an entirely new editorial article.

That moves GEO upstream into editorial intelligence.

I have written before about GEO-informed thought leadership and becoming the GEO journalist. This is the measurement consequence of the same idea.

If GEO can influence what we investigate before publication, then we should not judge it only by what happens after publication.

We should also judge whether it made the organisation learn something better in the first place.

There is a deeper opportunity here

For years, thought leadership has had a measurement problem partly because organisations try to measure it like marketing inventory.

A report goes out. Performance comes back. A dashboard determines whether it worked.

But good thought leadership is really an accumulating knowledge system.

You investigate something. You learn. You form an argument. People respond. Some parts travel. Some fail. The response creates the next question. The organisation becomes slightly more knowledgeable than it was before.

GEO may give us another sensor inside that system.

Not a perfect one, and not a replacement for brand research, audience response, PR, commercial evidence or editorial judgement. But a sensor capable of showing, at a surprisingly granular level, where an organisation is present in the answer environment, where it is missing, which evidence is travelling and which questions remain poorly answered.

The interesting opportunity is not to optimise everything for that sensor. It is to use the sensor to become better at deciding what deserves to be learned next. That is a very different ambition from ranking for prompts.

The feedback loop matters more than the score

I do not yet know how reliable this approach will be. That is partly why I want to test it.

The result at Accountex might be strong practitioner evidence that changes an answer. It might produce useful short-form video. It might improve an existing article. It might reveal that the visibility gap had been misdiagnosed. And the eventual GEO score might not move at all.

That would not automatically make the experiment a failure. If we learned something real about the audience, improved the quality of the organisation’s knowledge and made better publishing decisions, there is still value there.

If repeated interventions eventually produce observable changes in the answer environment as well, something more interesting begins to emerge: thought leadership starts gaining a feedback loop it has historically struggled to achieve. Not simply we published something good and hoped authority followed, but we observed a question, identified what seemed to be missing, learned something new, intervened deliberately and looked again. That is a much smaller loop, and perhaps a much more useful one.

GEO may not solve thought leadership’s attribution problem. But it might shorten the distance between a question, an intervention and what we can learn from the result.

For me, that is far more interesting than another argument that better content will somehow make brands more visible to AI.