AI works well when your content does too

Guest writer Tom Davis talks about the importance of good content and how Airgentic’s technology is helping to spearhead content governance strategies.

Many executives are talking about AI solutions.

One of the questions we’re hearing most: “We need AI, which AI tool should we use?”

There is an equally important question sitting beside it: when did anyone last look at what your website is actually saying?

These are not competing priorities, they are connected ones. And the organisations getting this right are looking at how their content interacts with AI tools. 

The way people find information has changed

Not long ago, the path to your business or organisation was fairly predictable. Someone would use a search engine like Google, find your website, land on your homepage and complete the task or find the information they need.

It doesn’t work that way today. 

People now find information through search engines, through on-site search on your website, through AI tools like ChatGPT, Gemini, and Copilot that provide answers without the person ever visiting your site.

There’s also voice assistants on their phone or smart speaker, and through assistive technologies that read content aloud for people with disability.

When it comes to people actually reaching a website, numbers are falling. This place more emphasis on creating a good experience to help people find what they need when they arrive.

The Australian Centre for AI in Marketing says, more than 70% of Google searches in Australia were producing zero clicks through to a landing page by early 2026.

Zero-click is where the user finds what they need in an AI overview. Software provider SEM Rush reports the volume of zero-click searches is on the rise in the US to almost 30% while other data quotes 60%. 

The common thread these channels have is they reward content that is clear, current and well-organised. 

When someone asks an AI tool a question, they don't get a list of links to choose from. 

They get a single answer drawn from whatever the AI could find and read.

If your organisation's content is out of date, buried in PDFs, poorly structured or written in jargon there's an opportunity to achieve clarity at scale. 

The “opportunity knocks” inside your content

In May 2026, Google published its first official guide on optimising content for AI search.

To people like me who sit at the intersection of communications and tech, I was pleased to see three messages which summed it up: content needs to be “helpful, reliable, and people-first.”

That’s from the organisation that dominates more than 90 percent of search traffic.

“Be sure that you're writing non-commodity content that your readers will find helpful and reliable,” It says.

Your organisation exists for humans at the end of the day.

And in a world where people are asking more and more questions in AI tools, relevance is more important than ever.

The next point that Google made is “Focus on what your users want, and avoid overdoing it.”

I talk often to my clients about making sure digital content and web pages follow a “need to know over nice to know structure.”

It's called the inverted pyramid.

Distracted customers aren’t the only ones who will get overwhelmed if they can’t find what they need.

Those tiny crawler bots from Google and AI tools which read and scan your content will too if it is bloated. They can then serve up “AI slop”. 

When did you last look?

Here is a question worth putting to your leadership team: when did anyone last review what your website is actually saying?

Not a redesign, not a fancy CX strategy, not a new content strategy. Just a simple review.

Is the information accurate? Is it current? Does it actually reflect how your organisation actually operates today?

The story tells the tale. Content was published, there was a restructure in the business. The web team was scaled down, staff moved on, policies changed, services evolved. A new head of digital came in. Yet, the website kept saying what it always said.

Old PDFs approved by the legal team five years ago stayed live.

This is what happens when organisations treat publishing as a one-time act rather than an ongoing responsibility.

Working with two organisations recently, an NGO and a government body, we used Airgentic's software to assess how their content was performing across search and AI channels.

What Airgentic revealed was a clear picture of content performance opportunities. Pages that were invisible to search engines. 

Duplicate content that doesn't help AI tools. 

Outdated PDFs that can be digitised. 

Both had the same underlying problem. 

Bloated content and a website structure not reflecting what they actually did.

The websites that had grown faster than anyone's ability to manage them.

In both cases, the information people and AI tools could find was not always the content that showed them in the best light.

This is exactly the kind of insight that Airgentic's technology surfaces quickly and clearly. Instead of guessing where the problems are, organisations can see them.

That changes the conversation with leadership from a vague concern about content to a specific, actionable plan.

With that data on the table, the path to being AI ready with voice activated search becomes much clearer.

What users experience when they arrive

The data from these organisations told a consistent story about what happens when content isn't governed.

People arrive overwhelmingly through search, often on their phones and they experience what we call is friction.

I had it in Melbourne when I couldn’t find the information to add the Myki transport card to my Google wallet.

It wasn’t on the webpage to “Buy a Myki card”.

A call to the helpful call centre solved the issue - 15 minutes later. I missed two trams, but I gave some feedback on the recorded call to the customer service person (a human) on the other end of the line.

Back to users in real life. They click on things that don't respond. They navigate to pages that don't answer their question. They scroll through dense, text-heavy content looking for a clear next step that isn't there. They give up and leave.

In the accessibility world that’s exclusion. 

I see it more often than not with information structured around how the organisation was built, not around the questions people were actually asking.

The business units have the pages designed for them, not the customer submitting the form or query. 

People in stressful situations,  trying to understand a process, check on a service, find out if they can get a grant to keep their lights on, deserved clearer answers than they were getting. 

Increasingly, people aren’t navigating the website at all. They're asking an AI. 

Why this is an ELT conversation, not just an IT or marketing conversation

Content governance is often treated as something the digital team handles.

We’ve heard it: “That’s a website problem.” “It’s a comms team problem.”

Something that sits below the strategic agenda.

The evidence says otherwise.

What we do know is that the same qualities that determine whether your content ranks in traditional search — clarity, structure, authority, freshness — are what AI tools will thrive on.

The intelligence of any AI tool is bounded by what it can find and trust.

A UK case study on enterprise AI

The UK Government Digital Service learned this directly while building an AI assistant for government services. Content in this instance is referred to as “data”.

“The underlying principle is simple: a Large Language Model (LLM) can only generate accurate answers if it is given the right information to work with,” says the UK Government Digital Service in Developing GOV.UK Chat: Our data science and AI engineering journey.

The content work came “before retrieval”, before the technology solution was built, not as an afterthought. In short, the content was vetted, structured and prioritised, first.

Principle one of the success criteria they used was “responses strictly follow retrieved GOV.UK content.”

“We filter GOV.UK content using metadata to prioritise authoritative, up-to-date sources and exclude any content containing personally identifiable information. We also use hierarchical semantic chunking, splitting pages into coherent sections while preserving the HTML header structure, so that content is well-structured and contextually meaningful before it reaches the retrieval stage,” the paper goes on to say.

The collaboration involved content designers and subject matter experts across GDS including data scientists working together on the solution. Humans in the loop working together. 

What good content governance actually requires

This is not about a website rebuild or a new content strategy.

Or saying wait to introduce an AI search tool. It is about getting key stakeholders together and answering basic questions.

Which underlines why “humans need to be in the loop” with each step. 

Who owns each part of your website? Not the web team. Who is the subject matter expert. Every section of your site should have a person responsible for keeping it accurate someone who knows when a policy has changed, a service has ended, or a process has been updated.

Without “content owners” nothing gets reviewed and nothing gets removed. There’s multiple risks here as we have covered. 

When was your content last checked? When should the regular review cycle be? For end of financial year reports, key milestones, announcements etc. What is needed to catch the majority of problems before they become embarrassing or costly. Most organisations don't have one.

Is your important information on web pages or buried in PDFs? PDFs are often invisible or misread by AI tools. Outdated PDFs are worse. They stay indexed and get surfaced long after they've been replaced.

If information matters, make it a web page that can be found, read and updated.

Is your content written for the person who needs it, not the person who wrote it? 

How many pages are using Plain language: short sentences, clear headings with the most important information first is not a style preference. There’s an ISO Standard for guidance for Plain Language.

Plain Language contributes to pages getting “ranked” by search engines and AI can surface it content accurately.

Does your team understand how people find information in today’s world? 

Many content decisions are made by people who assume visitors arrive at the homepage, read carefully and navigate logically.

Many overseas students for example are used to engaging with high-powered chat bots and “talking to do their searching” about a course.

The return on investment is real. Clarity opens content to everybody.

About Tom Davis  Tom Davis is a clear language advocate, award-winning journalist and founder of Black Sheep Communications, helping organisations turn complex ideas into clear, commercially valuable communication.

He works with government and major brands to embed clarity into content, strategy and digital transformation.

Tom is a member of Standards Australia’s Plain Language Working Group, contributing to the development of ISO 24495-1:2023 plain language standard and lectures in communications at the University of Queensland. 

He is also a member of the Localisation and Implementation Committee, established by the International Plain Language Federation, which facilitates the global adoption of the plain language standard.

Tom regularly speaks on Plain Language and AI and supports leading accessibility events such as the Perth Web Accessibility Conference.

Sources

https://dri.es/ai-and-the-great-cms-unbundling

Google Search Central Blog — "A new resource for optimizing for generative AI in Google Search" (May 15, 2026) https://developers.google.com/search/blog/2026/05/a-new-resource-for-optimizing

GOV.UK Blog — "Developing GOV.UK Chat: Our data science and AI engineering journey" (May 15, 2026) https://insidegovuk.blog.gov.uk/2026/05/15/developing-gov-uk-chat-our-data-science-and-ai-engineering-journey/

Semrush — "How to Win in a Zero-Click Search Market" https://www.semrush.com/blog/zero-click-searches/

ACAM / IT Brief Australia — "ACAM launches search course as zero-click AI rises" (April 14, 2026) https://itbrief.com.au/story/acam-launches-search-course-as-zero-click-ai-rises

Searchable — "What Is AEO? Answer Engine Optimisation Explained" (February 10, 2026) https://www.searchable.com/blog/what-is-aeo

Tracey Wilson, Substack — "We Don't Need Smarter Chatbots — We Need Smarter Content" (May 18, 2025) https://trackiew.substack.com/p/we-dont-need-smarter-chatbotswe-need