GB.png
May 21, 2026
AI-Ready Content: How Google Sees Government Websites (Part 2)
In our December 2025 GovChats, we sat down with Gabby Burke, an industry architect focused on health and human services for Google Public Sector. We spoke to her about how even as Generative AI changes the online content experience, the best practices of content strategy and getting your content seen remain largely the same. This is Part 2 of the conversation.
Content Strategy: The Big Picture
As Gen AI usage becomes a bigger part of the online experience, how do you think that agency website content should change? There are probably two parts to this: how can it better supply generative AI? But also how do you think content strategy should think about users who are still coming to the website?
I like the separation between how do we feed the AI and how do we continue to prioritize people. For the Gen AI side of the house, we've talked through some of these topics, but I would add on atomizing your content. Instead of having a 50-page policy document, for example, have a TL;DR — a more simple section.
Remember that LLMs are trained on language. LLMs do not think, but they kind of “think” the way that humans communicate in language, so in a lot of ways the benefits are going to be twofold: for the LLM and for the person, because I also don't want to read a 50-page PDF! Break down those large documents and really long, many-scroll web pages into smaller chunks with good headings. It's easy for the AI to understand how to answer specific questions, but it's also easy for a person to know what they're looking for and find that information quickly.
On the direct-to-user side, as you're thinking about your web content, it's good to start thinking about what people are doing, what action you want them to take, instead of what they should be reading and what research they should be doing about that action.
More folks are going to be jumping into your website knowing, here's the 10 steps that I have to do to renew my driver's license. They're going to be ready to take an action when they make it to your website. Assume that's the case and make that content more available to them. Don't expect they're going to be doing a deep dive. Instead, they're going to take that action because they already know what they're looking for.
It sounds like users will probably be a little bit more informed and prepared when they're coming to public websites now. As that changes how users find information and transact with the government, what metrics would you suggest agencies should be looking at and tracking?
We want to pivot toward a more outcome-focused evaluation approach. If you're using your web metrics for evaluating programs or understanding how your programs work, it's a good opportunity to think about what metrics are really important to you, because some metrics really are going to matter less.
For example: web traffic, page views, amount of time spent on page. We've used these for a long time as a placeholder of “we interacted successfully with this person,” but really it's not super aligned with the agency mission, right?
At the end of the day, regardless of which agency you are, you might not care that a person spent 20 minutes on your page. In fact, that might actually be a bad thing, depending on what they're trying to do. So I would say having outcome-focused metrics has always mattered, but we really haven't had a way to make that happen.
I think a lot of agencies already look at Google Trends and the kinds of things people are searching for. That will continue to be important. I think things like task completion rates, how many people actually went through and completed the task. Any of those outcome-focused KPIs will really matter.
A really impactful and interesting part of this evolution of the internet is going to be: how can we meet people as individuals? Not everybody wants to talk to a chatbot. Not everybody wants an AI Overview. Some people want to get on the phone. Some people want to be in person. Some people do want a chat experience. How can we more proactively ask for user feedback?
How can we use an opportunity to reach out and understand more qualitatively how people are interacting? How was the experience? Are they happy with it? Did they come and do what they wanted? Did they want to get on the phone and they were able to or not able to? It’s about customizing that for the particular use case.
Ask AI What It Wants
GOVChats: GenAI Search + Georgia
There's really no way to track how often your website's content is being sourced in AI summaries or even in something like ChatGPT. Do you think that will change and do you think it's necessarily even important to know that?
I don't work on the search side of the house, so I can't say with certainty, but if I had a crystal ball, I would anticipate that more data is going to become available over time from Google. I think Google Trends is a great example of this. Google has historically expanded the availability and transparency of search-related data as it becomes more available over time. So I do sort of predict there will be more data available down the road as this becomes more entrenched. This is very new, just in the past year or so. So as it becomes more widespread I do anticipate more data will become available.
Is it important? Maybe. I think it's important to understand whether your sites are interacting accurately with the LLM. But to get super tactical for a moment, there's already some other ways you can triage this type of information without having a specific service saying here's how many website hits we got.
I know it sounds kind of basic, but Overviews use Google's foundational model, Gemini. Do some informal testing. Drop your website links into gemini.google.com. Get some feedback. LLMs tend to be really good at understanding how LLMs want to see information. Get some feedback from Gemini and just see what comes up. Maybe it'll be helpful or give some pointers around areas that could be clarified.
Closing Thoughts
What really works with SEO, and now GEO, and with user experience, is having a holistic approach. What do you think would be the top best practices agencies should utilize to make sure their content is AI-ready, findable, and easy to interact with on their website?
- Embrace structured data. This is true for SEO, too. Make sure you're using metadata and markdown and all of the things that make the data really easy to find. That's going to be the top tactical thing.
- Take a user-centric approach to designing your content. You're probably already doing a lot of this, but consider: I'm a constituent. I'm coming to your website. What am I looking for? What questions am I going to ask? And structure your content accordingly. It's worth the extra work to do the restructuring so that folks can have a people-first interaction with your website, and that'll help the AI as well.
- Aggressively prioritize your authority. Make sure you're citing authors and sources where it's appropriate. State when content was reviewed or updated. Make sure it's clear to the LLM what the newest thing is because that'll really help make sure it gets bumped up in that trustworthiness metric.
- Be proactive. Make sure that you're addressing things head-on. If you need to say on a website in a little box we heard about this scam, here's the parameters and what you need to do instead, don't be afraid to have that proactive content. That really helps.
Optimize your website for action. Assume people are coming to your website more informed. Maybe not today, but down the road, people are going to be looking for actions from websites instead of just information. Make that as easy as possible for constituents so they can do what they need to do on your page really fast.
This interview has been abridged and edited for clarity.
Related to:
