Findability and AI for councils

Residents look up the same things every week: bin days, rates, dog registration, consents. More and more, the first answer they get comes from Google, ChatGPT or Claude, not from the council. I check what they're told, where each answer comes from and who needs to fix it, and give independent advice on council AI projects.

I work with councils anywhere in New Zealand, drawing on the findability work I led at MBIE.

Where councils feel it

Residents ask the same things every week. Here's what I check for each question, and what you get back.

A resident asks Google

“When's my rubbish collected over Christmas?”

What I check
Whether Google's AI Overview, ChatGPT, Claude and your own site search give this year's dates, and which page each answer comes from.
What you get
The old page behind any wrong answer, and who updates or redirects it.

A resident asks ChatGPT

“How much is dog registration this year?”

What I check
Whether the answer quotes this year's fee, last year's, or a neighbouring council's.
What you get
The pages AI tools draw on, and what to change so the current fee is the one they find.

A resident follows an old link

“Who do I call about a water leak on my street?”

What I check
Whether old links and old names still point to the council after services moved to a new organisation.
What you get
The broken links and out-of-date pages, with the owner of each fix: the council or the new organisation.

A resident searches your site

“building consent for a deck”

What I check
Whether site search and Google find the right step across the main site, the consents portal and the district plan.
What you get
A fix list naming who acts: web team, platform owner, comms, customer services or your CMS vendor.

A staff member asks

“Can I paste this complaint into ChatGPT to draft a reply?”

What I check
Which tasks staff already use AI for, and what privacy and LGOIMA mean for what goes into a prompt.
What you get
A safe-use policy people can follow, and the tasks where AI is worth using.
ChatGPT on the cost of registering a desexed dog with Wellington City Council: the sources it used are Petme and a pet-site fee calculator, and a harness ad sits under the answer
ChatGPT gets Wellington's dog fee right, from a pet site, with an ad underneath.

A first answer review: what residents are told about your council

It's a findability audit built around residents' questions. I take the questions people actually ask, in their words, and check the first answer they get from Google, AI assistants and your own site search. Then I trace each wrong or out-of-date answer back to where it came from.

Wellington City Council's dog fees table: desexed dog $148.00, or $222.00 if paid on or after 1 August
The council's own figures are in a collapsed table on its fees page.

That's rarely just the main website. Most councils also run a consultation site, the district plan and maps, online services, and sites for libraries, venues and pools, and share information with a water organisation, civil defence and the regional tourism site. Answers come from outside too: a neighbouring council, an old PDF, a news story. I'm independent of your CMS and platform vendors and look across all of them. The review covers:

  • What residents are told. Rates, rubbish and recycling, consents, dog registration and other high-volume services, with each answer checked against your own information.
  • Where the answers come from. Which of your sites, platforms and outside sources each answer relies on, including ones you didn't expect.
  • Search visibility. Whether the right page comes up in Google for those questions, and the technical SEO basics behind it.
  • Broken inbound links. Links from Wikipedia, community and government sites that no longer work, and old addresses left over from earlier restructures. If services are shared or councils combine, every old link needs a new home.
  • AI crawler access. Which AI tools can read your sites, so blocking them is a decision, not an accident.

Each answer in the report is set out like this:

Answer check Wellington City Council Checked 27 Oct 2026

  • "How much to register a desexed dog?"ChatGPT, citing Petme and a pet-site calculator

    Unofficial source

    ChatGPT said$148, or $222 if paid late. Right, but taken from a pet social site, with a harness ad underneath.

    The council says$148, or $222 on or after 1 August, in a collapsed table on its fees page.

    Fix: put the main fees in the page text, not only in a collapsed table.

The dog fee question from the two screenshots above, as it appears in the report.

You receive a written report and a fix list that names who acts: your web team, the owner of each platform, comms, customer services, IT, your CMS vendor, or a council organisation such as your water provider. More on the findability audit.

Advice on council AI projects

Independent, vendor-neutral help when a council is deciding what to do with AI:

  • Staff use of AI. Which tasks are worth it, a safe-use policy people can follow, and what privacy and LGOIMA mean for what goes into a prompt.
  • A chatbot or AI search for residents. Whether it's worth doing, what it should and shouldn't answer, and testing it with residents' real questions before launch.
  • Choosing a vendor. Requirements you can take to market, and a comparison on your own content rather than the vendor's demo.

Councils aren't bound by the central government AI guidance, but it's a sound reference, and I work with it. I advise, specify and test; your team or your vendor builds. More on AI project consulting.

Try the bin day test

You can see this for yourself in a few minutes. Take five questions residents often ask the council, such as which day the bins go out, how to register a dog or when rates are due, and ask them in Google, ChatGPT and Claude. Compare each answer with your own information. Look for answers that are wrong or out of date, and for sources you didn't expect. If you find problems, a first answer review works out why and what to fix first.

Further reading

Common questions

Isn't this something our website vendor does?

Your vendor looks after the CMS, and that matters. The review looks at the answers residents get, which come from several council platforms and outside sources as well as the main site. I'm independent of all of them, and the fix list includes work for your vendor where it's theirs to do.

Can you work across several councils in our region?

Yes. AI tools often mix up neighbouring councils, especially where services are shared or names are similar. A regional review tests the same questions for each council, shows where the answers cross over, and can be shared between the councils or run through a shared services group.

Do you work with councils outside Wellington?

Yes. I'm based in Wellington and most of the work is done remotely, with a start call and a walk-through by video. I can come on site when it helps.

How does procurement work?

We start with a short call about what you need. I then send a written quote with the scope and a fixed price, agreed before any work starts. A review or a single piece of AI work is sized to fit within the single-quote limit in most councils' procurement policies, and your procurement team confirms the purchasing route and any supplier requirements.

We're a small council with a small web team. Is this for us?

Yes. The fix list names who acts, whether that's your web person, the owner of a platform, IT or your CMS vendor, and puts the few things that matter most first. Small teams get the most from knowing what not to spend time on.

Can you look at our staff use of Copilot, ChatGPT or Claude?

Yes. That's often where council AI work starts: which tasks are worth it, a safe-use policy people can follow, what goes into a prompt, and how to check what the tools produce before anyone relies on it.

Talk about your council

Send a few lines through the contact form or call (027) 450 5720.

Send a message