Generative Engine Optimisation (GEO) Tips for Newsrooms

A certain amount of irony was detected in the creation of this image

Before I start, let’s talk about a critical point. The way LLMs generate answers might differ from Google’s crawling and indexing behaviour, but the fundamentals of optimising for both experiences are pretty much the same. The difference between GEO and SEO, if there really is one, is perhaps your priority list of what you want to tackle first and the tools you might use to measure success - which is, itself, currently up for debate due to the probabilistic nature of LLMs.

A publisher website that is not truly audience-focused, does not demonstrate authority on a topic, is hard to crawl and does not shout about the credibility of its authors IS going to have the same visibility issues no matter what platform.

Nonetheless, I thought it would be interesting to tackle the subject of AI optimisation for newsrooms through the GEO lens, if nothing else, to demonstrate the similarities with SEO.

And also let’s not forget that currently, the average referral rate from these LLMs are on average only 1% - and sometimes even lower. 

Let’s repeat that: 1%. 

But I get it - you’re being asked to deliver a GEO plan for your team, so let’s crack on. 

The current landscape of GEO

We all know search is changing fast. Readers still Google, of course — but more and more are getting information through AI-generated answers that summarise the web and cite sources selectively.

For newsrooms, that creates a new challenge (and opportunity): how do we make our reporting the kind that LLMs choose to reference? 

That’s where Generative Engine Optimisation (GEO) comes in.

GEO, like SEO,  isn’t about “tricking” AI. It’s about making your journalism:

• easier for machines to understand and extract

• easier to quote accurately

• more clearly attributable

• more likely to be selected as a primary source
  

So that’s the preamble, let’s get to the meat of it.

What GEO Actually Means (in Plain English)

Generative engines don’t rank ten blue links the same way traditional search does. Instead, they try to:

1. interpret the question,

2. pull in sources that look reliable and relevant,

3. produce a single narrative answer (often with citations).
   

So GEO is about increasing the odds that your work is:

• ingested (accessible to crawlers),

• understood (clear structure and entities),

• trusted (signals of authority and transparency),

• citable (clean facts, numbers, dates, and quotes).
  

I might also add at this point that although the underlying technology behind LLMs is equally fascinating and complex, the way they crawl and extract information is quite rudimentary compared to Google which has been in the business of information retrieval for longer. This has opened wide the market for some very questionable (and very old) tactics to optimise for LLMs at the risk of, let’s be frank, pissing off Google.

No one wants to receive a penalty from Google, particularly because LLMs reference either  Google or Bing results (RAG/Grounding). If your visibility drops suddenly in Google, expect the same result in LLMs. Following that logic, boosting your visibility in Google and Bing will boost your visibility for LLMs - so hire a good SEO to get the job done!

So here are my GEO recommendations for newsrooms if they are aiming to boost their visibility in LLMs.

Step 1: Follow the 5 Whys

LLMs love content that answers quickly, and journalists have been doing this for yonks with the 5 Whys and inverted pyramid model:

• Who is the story about? 

• What happened? 

• When did it happen? 

• Where did it take place?  

• Why did the story occur?  

• How did the events in the story come about? 
  

Example “Southwark Council approved a 12-month pilot on 16 December 2025 restricting traffic on Walworth Road, London, aiming to reduce air pollution near schools.”

That one sentence gives a generative engine exactly what it needs: entity, action, date, place, purpose.

I bet you might be frowning a little at this example isn’t exactly ground-shattering. It’s something that news journalists write every hour of their lives (well, not literally about Southwark because that would be odd) - but it does highlight the fact that waffly, garrulous copy is just off-limits when it comes to GEO. 

In many respects, good journalism IS good GEO, but just remember that LLMs are always looking for concise passages that e.g. answers a question or explains what an entity is. This is why this horrible term called 'chunking' has emerged. My take on this is 'chunking' lark is that it might suit forms of technical writing but is definitely NOT applicable to all forms of news journalism.

Getting cited in an LLM by following strict GEO guidelines might not be the primary goal of your article. You may, instead be writing a first-person story that is so brilliant that. although it MIGHT be summarised by an LLM, the reader experiences complete FOMO and is compelled to download your app to read more. Like a radio or podcast presenter discussing the news - the content could be easily summarised, but that is not why listeners are loyal to these personalities everyday.

Step 2: Use Specific Entities (Not Vague References)

Generative systems work best with clear “named entities”:

• people (full names and roles)

• organisations (full official names)

• locations (specific sites, not “the area”)

• laws/policies (official titles)

• datasets (source name, release date)
  

Avoid: “the minister”, “the watchdog”, “the local authority”

Instead: “the Home Secretary”, “the Competition and Markets Authority (CMA)”, “Bristol City Council".

This reduces misattribution and increases quote-ability. Again, your friendly SEO lead would have already spoken to you about the importance of including the Prime Minister’s full name rather than surname in a headline (for instance)  - which I still see as a battleground for some SEOs out there.

Step 3: Add a Summary Box 

You know all about this, it’s pretty much what EVERY newsroom did when ChatGPT first entered our (mainstream) lives in 2023.

This is simply about adding a summarised list at the top of your articles, brief lines that might include: 

Key points

• What happened

• The latest confirmed numbers

• The official source(s)

• Next milestone/date
  

The theory being It helps humans scan, and it gives AI clean chunks to summarise without mangling nuance.

The thing is if you do it for EVERY article you publish it can become pretty tedious on the eye.

But what I do know (psst…from my newsroom peers in the industry) is that these summaries can be pretty good for audience retention in the article. Readers don’t read the summaries and then disappear, they tend to stick around and engage with the article for longer.

So, perhaps, summaries are valuable, but I would just be cautious about rolling it out across your entire publishing remit. 

Step 4: LLMs Love a Good Table...oh, and Headers

The reason why LLMs love a good comparison table is because you are basically doing all the hard work of comparative analysis for them. So yes. LLMs love a good table in the same way they love a good comparison page. In fact you don't NEED an actual table - it could simply be comparing product a, to b and c in the page copy - but sometimes it's just easier for readers to parse if there is a visual.

Then there is page structure. Make it really easy for LLMs to understand how information is being ordered by maintaining a clear hierarchy (H2 for themes, H3 for specifics, bullets for clarity). Again this is not going to suit EVERY news article format (opinion or interviews for instance) but you get the idea.

Step 5: Make Facts Easy to Verify (and Hard to Misquote)

Generative engines tend to prefer content that looks verifiable.

Do more of:

• precise dates and times

• exact figures with units (and what they refer to)

• direct links to primary documents

• clear separation between fact and analysis
  

Example phrasing that reduces AI mistakes:

• “According to ONS data published on [date]…”

• “The council report states… (link)”

• “Police said at [time] that [number] people were affected.”
  

This is why PR and marketing teams might need to review copy about their brand that celebrates what the brand has achieved - but doesn't have the data or case studies to back it up. Working with PR and Marketing teams is important with AI optimisation because LLMs often references text that might live on a news brand’s corporate website - far away from the prying hands of an Editor who could quickly update it. So make sure all teams are aligned because in GEO, you are only as credible as those (often outdated) pages say about you and what OTHERS say about you.

Step 6: Cite Primary Sources Like You Mean It

AI answers often privilege sources that appear to anchor claims in documents.

When possible, link to:

• legislation and statutory instruments

• government announcements (GOV.UK)

• local authority meeting papers

• regulators (Ofcom, CMA, FCA, ICO etc.)

• court judgments / official transcripts

• ONS datasets / NHS guidance
  

Even if you are quoting from these sources, actually linking matters.

Step 7: Build “Topic Authority” With Structured Coverage

Generative engines look for patterns of expertise.

Newsrooms can create authority by:

• maintaining an explainer hub for complex topics (e.g., asylum policy, NHS waiting lists, rail strikes)

• using consistent tagging and internal linking

• updating evergreen explainers when stories move
  

I’ve already written about the importance of internal linking for newsrooms and for GEO, just like SEO, you need to make it exceptionally easy for the LLM to crawl your website. That’s another big topic that I’ll write about soon, but in the meantime a good Technical SEO can help you achieve this. 

Incidentally, LLM visibility of evergreens starts to diminish after three months, so make sure you have a system in place to update them. Track your most linked and cited pages and keep them fresh. Incidentally, LLMs like articles that are comprehensive, clear, and easy to read. This usually suits the explainer format but also deep dives, investigations, profiles.

Step 8:  Use Plain English.

AI systems often struggle with obscure official language and acronyms (who doesn't?) so help them.

Add short context lines:

• “A Unitary Authority is…”

• “The Section 114 notice means…”

• “The King’s Speech sets out…”
  

This reduces the risk of incorrect summaries and increases your value as a clarifying source.

Step 9: Use Schema and Structured Layouts

Even in GEO, technical accessibility is foundational. If AI crawlers can’t reliably parse your page, you’re invisible.

Newsroom basics that help:

• Article pages that render server-side or are crawlable without heavy JS

• descriptive titles, canonical URLs, proper indexing controls

• NewsArticle / Article structured data with:

  • headline, datePublished/dateModified

  • author and publisher

  • mainEntityOfPage

  • paywall markup if relevant
    

If you run a paywall, clear signalling can help systems understand what’s accessible and what isn’t.

There's a lot more you can do with schema but recent studies suggest that LLMs seem to care more about whether the information is structured (via headings) than whether it’s technically marked up.

This is just a mere tip-toe into Technical SEO and again, a good Technical SEO (hint, hint) will be able to assist you with this work.

Step 10: Be Clear About Ownership, Corrections, and Updates

Trust signals really matter in GEO, as they do in good ‘ol SEO (remember EEAT anyone?) 

Make sure:

• author pages exist and show expertise

• corrections policy is easy to find

• updates are timestamped (“Updated 14:32”)

• original reporting is labelled as such

• syndicated copy is identified
  

This helps LLMs understand how trusted you are as a publisher. A game I like playing in LLMs is asking them ‘Is X newsroom a credible news source?' Also useful, for individual journalists. Even though LLMs never reveal their inner workings, although they might appear to, it’s still a useful exercise. 

Step 11: Community Matters

Currently, user generated content  and community mentions, especially on platforms like Reddit and YouTube, are major drivers of AI citations.

However, faced with Zero-Click, optimising for platforms outside of Google isn’t really just an AI Visibility play - it’s a search EVERYWHERE play. Own your audience, everywhere. 

Also, be careful in those subreddits. The communities can be brutal with brands that don't respect the guidelines.

Step 12: Should your newsroom go listicle mad? (Er, Not really)

Because of the rudimentary way LLMs assign authority to sources, this has led many publishers to write (or assign a third party to write) multiple ‘Best of’ articles on a topic with their brand at the top. The idea being that LLMs love a good ‘Best of’ article and will assign credibility to the brands in the list. This kind of practice has always given SEO a bad name and I don’t recommend it as a primary directive, especially when audience trust is so important to newsrooms. Proceed with extreme caution.

Step 13: Don’t Optimise Yourself Into Blandness

A fair worry: if we write for machines, do we lose our voice?

I think this is the same principle for SEO writing when we say ‘write for humans’ - yes write for readers, but remember to take a structured approach with a clear and explicit intention.. 

A Simple GEO Checklist for Editors

Before publishing, always ask:

• Does the first paragraph clearly state who/what/when/where/why?

• Are key entities named precisely (people, orgs, places)?

• Are important figures dated and sourced?

• Is there a clear hierarchy to the page? - headers, sub headers, bullets.

• Is there a short Summary section? (optional and worth testing)

• Are primary documents linked?

• Is this article part of a wider internal linking ecosystem?

  
  

The Big Takeaway (and I don’t mean noodles): GEO Rewards Clarity and Credibility

Generative engines are trying to answer questions quickly. UK newsrooms win when we make it easy for systems to:

• understand the facts,

• attribute them correctly,

• and cite the right source.
  

In other words, GEO is like SEO in this regard - the success of both depend upon excellent digital journalism practice - with a structured approach to writing. 

Of course, optimising content so it appears on a third party platform is really just one-side of the coin. The other should be focused on making your own website or app as captivating as possible and I have lots of ideas for engaging audiences off-LLM right here.