Becoming the GEO journalist

The next stage of GEO is not just answering questions better. It is deciding whether an important question needs a better page, better use of evidence that already exists, or genuinely new reporting.

One version of GEO turns the content team into optimisers.

Find the questions appearing in AI search, audit the pages meant to answer them, and rewrite what’s weak — tightening the structure, adding citations, making the existing material easier for machines to retrieve.

All useful, and it’s the version I’ve argued is optimising the wrong layer when it’s the only version running: it can only ever improve the answer an organisation already has, and says nothing about whether that answer was any good to begin with.

But I think there is another role emerging alongside it: the GEO journalist.

I originally framed that role mostly around a question: why don’t we have a good answer yet? If the evidence doesn’t exist inside the organisation, somebody has to go and get it.

That still matters. But I think the model needs one more step.

Sometimes the organisation already has the useful evidence. It just lives in the wrong place: a customer story, a research report, a webinar transcript, an event interview, a video, a sales conversation or a forgotten article. The job is not to create another asset. It is to recognise that this existing evidence could make an important answer materially better, route it into the right place and republish the improved answer.

That makes GEO an editorial renewal problem as well as an optimisation and reporting problem.

Three editorial routes: renew, route or report

A weak answer can mean several different things, and the distinction matters because each should trigger a different kind of work.

1. Renew the existing answer

The organisation already knows enough and the relevant page is broadly the right destination, but the answer is stale, poorly structured, incomplete or too generic.

The intervention is editorial renewal: update the page, make the answer clearer, add current context and improve its structure and sourcing.

2. Route existing evidence into the answer

The evidence exists, but it is stranded somewhere else.

A useful customer quote is buried in a case study. Research that explains the problem sits in a PDF. A practitioner interview contains a better explanation than the current product or advice page. A video answers the question clearly but the written ecosystem never uses it.

The intervention is not another net-new asset. It is to connect the evidence to the answer that needs it.

3. Report what the organisation genuinely does not know

The current answer is weak because the organisation lacks the evidence itself.

It may know that something happens, but not why. It may have a statistic but no explanation. It may have a product claim but no customer proof. It may have a broad point of view but no first-hand account showing how the issue plays out in practice.

No amount of rewriting or internal linking fixes that. That is where reporting starts, and the useful editorial question therefore becomes:

Do we need to improve the answer, reconnect evidence we already possess, or go and learn something new?

That is more precise than treating every weak AI answer as either an SEO problem or a commission for another article.

GEO as a reporting and renewal brief

Imagine an AI visibility review identifies a cluster of questions that matter to your audience.

Some are already answered well. Leave them alone or strengthen the connections around them.

Some are answered in the right place but need renewal.

Some have good evidence elsewhere in the organisation that should be routed into the answer.

Some expose a genuine knowledge gap. That’s where the workflow starts to look less like optimisation and more like journalism.

You can ask:

• What do we actually know about this?
• What evidence supports the answer?
• Is stronger evidence already sitting somewhere else in the organisation?
• Where is our answer generic because everybody is drawing from the same public information?
• Who has first-hand experience we haven’t captured?
• What would we need to investigate before we could say something genuinely useful?

Only the final group automatically requires new reporting.

Interview customers, talk to practitioners, question subject-matter experts. Analyse existing research and commission new research where the gap warrants it. Capture disagreements rather than smoothing them away.

The GEO analysis tells you where stronger knowledge may be valuable. The Evidence Engine helps capture, connect and maintain that knowledge — whether it is newly acquired or rediscovered in material the organisation already owns.

What this looks like in practice

A more complete operating loop is becoming clearer to me:

audience question → current answer → evidence map → renew / route / report → stronger answer → republish and connect → measurement

The important step is the evidence map.

Before commissioning another article, ask two questions:

1. Is the source material already good enough to support the answer?
2. If it is, is that material actually present in the place where the audience or AI system needs the answer?

If the evidence is strong and already in the right place, improve the content only where needed. If it is strong but stranded elsewhere, route it into the answer. If the evidence is genuinely missing, identify what is missing before anyone starts drafting.

That missing evidence can then shape questions in work that may already be happening: a customer interview, a video shoot, an event, a webinar, a roundtable, a research project or an expert conversation.

Those activities become opportunities to acquire knowledge the organisation does not yet have. But the archive becomes part of the production system too: existing research, customer stories and interviews become inputs that can improve current answers instead of remaining frozen inside the asset they were originally created for.

Reuse is not the same as repurposing

Content teams have talked about repurposing for years. Usually that means turning one asset into several formats: the webinar becomes a blog post, clips and social posts; the research report becomes an infographic and an email sequence.

That can be useful, but this is slightly different. The question is not what else can we make from this asset? It is where else does this evidence improve an answer we already need to give?

That might mean a statistic from a research programme strengthening an existing guide. A customer story adding proof to a product or use-case page. A practitioner explanation from an event updating an evergreen article. A video transcript revealing a better answer to a question already visible in AI search.

The source asset can remain intact. The evidence travels.

That is closer to knowledge management than traditional content repurposing, and it is one reason I think GEO creates an interesting editorial job. Someone has to recognise that the useful unit is often not the article, video or report. It is the evidence inside it.

A worked example: from statistic to explanation

Take a question such as:

Why do professionals keep doing work outside scope without charging for it?

Imagine the organisation already has quantitative evidence showing that the behaviour is common. That proves the problem exists.

It does not necessarily explain the mechanism behind it. You could write an article immediately and fill the gap with plausible explanations:

• fear of damaging the client relationship
• work arriving in fragments that feel too small to invoice
• unclear boundaries between service and extra work
• fixed-fee models making additional charging awkward
• a professional culture that frames extra help as part of the relationship
• uncertainty about when helping becomes scope creep

All of those may be reasonable hypotheses. But they are still hypotheses.

Before commissioning new interviews, though, check whether the organisation has already asked customers or practitioners something close to this elsewhere. There may already be a transcript or customer story that explains the behaviour in first-hand terms. If so, the first intervention is evidence routing: bring that explanation into the current answer with the right provenance and context.

If it does not exist, the reporting question is different:

Tell me about the last time a client asked you to do something you knew was outside the agreed scope, but you didn’t charge them. What did they ask for, and what stopped you charging?

That question can produce something the content system did not possess: a first-hand explanation of the decision itself.

A follow-up might be:

What would have needed to be different for you to feel comfortable charging for it?

Now the eventual written answer can move beyond what happens into why it happens, supported by evidence rather than confident inference.

Search intelligence can tell you where an audience has a question. Editorial judgement can tell you whether the organisation already has enough knowledge to answer it properly, whether that knowledge is in the wrong place, or whether reporting needs to close the gap.

The GEO journalist sits between demand, archive and evidence

That gives the role a clearer definition. The GEO journalist sits between demand intelligence, organisational knowledge and evidence creation. The technical GEO layer may identify:

• recurring prompts
• weak visibility
• missing or fragmented answers
• competitors being cited where you are absent
• themes around which an audience repeatedly asks questions

The editorial layer then asks:

• Which of these gaps actually matter?
• Which are simply content-structure problems?
• Which existing pages should be renewed rather than replaced?
• Which useful evidence is stranded in another asset?
• Which reveal that we genuinely do not have enough evidence yet?
• What would we need to learn before writing the answer?
• Who could credibly help us learn it?
• What question would get beyond the rehearsed response?

Then the editorial intervention follows the evidence rather than defaulting to a format.

This is not a replacement for technical GEO work. It is an upstream and sideways extension of it: upstream into reporting, and sideways into the organisation’s existing knowledge base.

The publishing system still matters

There is a practical constraint here that is easy to miss when describing elegant AI-enabled workflows.

The analysis layer can become very fast. AI can help compare a visibility question against a large content library, surface relevant customer evidence, suggest which pages could benefit and draft a proposed update.

But many organisations still publish through conventional CMS workflows. A WordPress site, for example, may still require somebody to open the page, apply the changes, check formatting, manage approvals and publish manually.

That does not invalidate the model. It changes the workflow design.

If publishing is the expensive step, the intelligence layer should produce a compact content renewal brief rather than a stream of vague recommendations. For each proposed change, capture:

• the audience or GEO question
• the existing target page
• the evidence to add or update
• where that evidence came from
• the specific editorial change proposed
• whether the page should be refreshed, expanded, connected or left alone
• who needs to approve or publish it
• what should be re-measured afterwards

That creates a bridge between an AI-assisted evidence system and a manual CMS without pretending the CMS is AI-native.

It also gives the team a way to batch and prioritise updates. Five high-value changes to existing pages may be more useful than five new articles, especially when publishing capacity is constrained.

Not every gap deserves a reporter — or a rewrite

An AI visibility review can surface dozens of weak or missing answers in an afternoon. Reporting on even one of them properly — real interviews, real verification, an editor deciding what the evidence actually supports — takes days, sometimes weeks.

Treating every gap as a reporting assignment would recreate exactly the problem I’ve argued content teams already have: commissioning by default because a system generated a plausible reason to, rather than because the question is actually worth the resourcing.

But treating every gap as a rewrite creates another kind of waste. Sometimes the existing page is not the problem. Sometimes the organisation has already produced a strong explanation elsewhere and simply failed to connect it.

The gate should therefore be: why this gap, why us, why now, and what is the smallest credible intervention?

That may be a paragraph update. It may be routing a customer example into an existing page. It may be a new interview question at an event already happening. Occasionally it may justify a full research or reporting commission.

A visibility tool can tell you a question is weakly answered. It cannot tell you which editorial intervention is worth the cost. That’s still an editorial call.

None of this works without the review that surfaces the gaps in the first place, which is a distinct piece of work from the intervention itself — an AI Visibility Audit is what that review looks like as a standalone engagement.

The opportunity is bigger than events

Events make the model easy to see because lots of useful people happen to be in one place.

Instead of arriving with a brief that says make some event content, an editorial team could arrive knowing the questions it wants to investigate. Interviews and video become ways of gathering evidence against those questions. Unexpected answers become new signals rather than inconvenient deviations from the content plan.

But the same idea applies almost anywhere an organisation is already talking to people.

A webinar can be designed partly around questions where the evidence is weak. A podcast interview can test an emerging argument. A roundtable can deliberately explore disagreement instead of resolving it too fast.

Customer interviews capture language and experience that strengthen a wider knowledge territory; sales and support conversations expose the questions customers keep asking that the published knowledge base barely addresses.

Even an executive interview becomes a chance to turn tacit expertise into attributable evidence.

Even a video shoot changes slightly. The goal isn’t only to obtain a polished film. It is to capture useful, self-contained expert answers that can also become transcripts, quotations, written answers, sales material and source evidence for future work.

The activity becomes a signal-gathering opportunity before it becomes a format — designed signal gathering is the fuller workflow version of the same move, for events specifically.

For each important audience question:

1. What do we already know?
2. Where does that evidence currently live?
3. Is the existing answer using the strongest evidence available?
4. What is still assumption?
5. What first-hand evidence would make the answer materially stronger?
6. Is someone we are already interviewing qualified to provide it?
7. What question is most likely to produce a specific, usable answer?

Then change only the interview questions that genuinely need changing.

The aim is not to make every interview serve GEO. It is to stop wasting both kinds of access: access to people who could answer questions the organisation needs answered, and access to evidence the organisation already paid to create but never reused intelligently.

A different content loop

The conventional event-content loop often looks something like:

event → interview → video/article/social posts → distribution

A GEO-informed editorial loop can be broader:

audience questions → visibility gaps → evidence map → renew / route / report → verified evidence → connected answers → multiple formats where useful → measurement → new questions

That distinction matters.

The first workflow begins with an occasion or asset and asks what content can be extracted from it. The second begins with something the organisation needs to understand or answer better and asks what the smallest useful intervention is.

Sometimes that intervention is new evidence. Sometimes it is a better use of evidence that already exists.

It means GEO doesn’t have to push organisations towards producing even more generic content. It could help them publish less, reuse better and report more selectively.

First-party evidence is not a magic ranking trick

There is a temptation to turn this into another technical GEO claim: AI wants evidence, therefore original reporting will rank. That is too neat. I would not claim that acquiring a customer quote or conducting an interview mechanically causes an AI system to cite your page.

The stronger argument is editorial. First-party evidence can make an answer:

• more specific
• more differentiated
• more attributable
• harder to reproduce from generic public information
• more useful to a reader trying to understand what actually happens in practice

The same is true when that first-party evidence was captured six months ago in another format and is only now being connected to the right answer.

That gives search and AI systems better source material to work with, but the visibility effect should be measured rather than assumed. The work is worthwhile because it improves the underlying knowledge asset first. GEO then gives us a new way to decide where that investment may matter most.

Original reporting becomes more valuable when production gets cheaper

Generative AI is extremely good at reformulating information that already exists. That makes another competent summary progressively less defensible as an editorial asset. A customer explaining what actually happened inside their business is different.

So is proprietary research. Practitioner experience. A useful disagreement between experts. A surprising observation from a room full of customers. A claim that can be traced to somebody who has actually done the work.

These things existed long before GEO. Journalism, qualitative research, voice-of-customer work and good content marketing have always depended on them.

What’s new is the possibility of using AI visibility as one input into deciding where to point reporting effort and where to reactivate existing evidence.

That creates a useful division of labour. AI can help identify patterns, connect transcripts, compare existing material, expose gaps, find relevant evidence in an archive and draft structured renewal briefs.

Humans can decide which questions matter, recognise when an answer is suspiciously generic, judge whether two pieces of evidence really belong together, ask the follow-up question, notice the unexpected comment and decide whether the evidence actually supports the story.

Where I’ve seen a version of this work

I don’t have a case study that started with an AI visibility gap and ended in this full renew-route-report loop — the framing here is new, so I should say that plainly rather than let a real example imply otherwise.

What I do have is Accountex 2026, which is the closest precedent: Hidden Hours research launched into a room full of the practitioners it described, and what mattered wasn’t the data alone. It was going in with questions still open and treating the room’s reaction as evidence rather than as an occasion to cover.

That’s the reporting instinct this piece is arguing for, running before the GEO framing existed to name it. The new part is using AI visibility and the existing content estate together to decide what happens next.

The next test is more operational: start with a defined set of audience questions, inspect the current published answers, map what existing research and customer evidence can already strengthen, then use planned practitioner interviews to acquire only the evidence that is genuinely missing.

The output should not automatically be a new article. It could be an updated evergreen page, an existing guide republished with stronger evidence, a customer example routed into an MTD or product answer, a better video question, or a genuinely new piece of reporting.

That is much closer to a real GEO journalism workflow than simply optimising finished articles afterwards.

Don’t let GEO dictate the reporting

There is an obvious failure mode here. If every interview question is derived from what an AI visibility tool says people are asking, you’ve just created another optimisation machine. Reporting needs room for surprise.

A useful brief might deliberately combine directed investigation with open listening. Go in with questions worth answering, but leave enough space for somebody to tell you that you’ve been asking the wrong question.

The same applies to content renewal. A visibility gap should not become an excuse to force every piece of available evidence into a page. Evidence belongs only where it actually improves the answer.

That is part of the value of keeping editorial judgement in the loop.

The strongest case against this

The honest objection isn’t that the idea is wrong. It’s that most content teams don’t have anyone who can actually do all of it well.

Original reporting is a specific, trained skill: knowing which follow-up question gets past the rehearsed answer, when a source is worth pressing and when to let a contradiction stand instead of tidying it away, what’s usable on the record versus what needs corroborating before it can carry an argument.

Evidence routing also needs judgement. A model may find a superficially relevant quote or statistic, but somebody still has to decide whether it is current, attributable, representative and appropriate for the claim being made.

And traditional publishing systems can make a theoretically elegant workflow operationally slow if every page update requires multiple manual steps and approvals.

Those are real constraints on adoption, not reasons the model is wrong. They are also arguments for why editorial jobs are moving upstream rather than disappearing. The valuable work moves towards judgement: what matters, what evidence belongs, what needs to be learned, what should be updated and what should be left alone.

Maybe this is one of the new editorial jobs

I don’t know whether “GEO journalist” will ever be a real job title. It probably doesn’t need to be.

But the function makes more sense to me now.

Someone sits between AI visibility, the published content estate, organisational knowledge and original reporting. They don’t just ask how to make existing content more retrievable.

They ask:

• What important question are we weak on?
• Is the answer itself weak, or just the page?
• Do we already possess stronger evidence somewhere else?
• Can an existing page be renewed instead of creating another asset?
• What do we genuinely need to go and learn?

Then they find the smallest credible intervention and make sure the result travels back into the information environment.

The question I am now most interested in testing has expanded:

Can GEO intelligence influence both what an organisation goes out and learns and what existing knowledge it brings back to life?

If the answer is yes, GEO stops being only a distribution and optimisation discipline. It becomes one input into editorial investigation, knowledge reuse and content renewal itself.

In that version of GEO, the editor doesn’t become less important because machines can produce more content. The editor gets sent further upstream — towards the questions and evidence — and further across the organisation, towards the useful knowledge that already exists but has not yet found its best home.