For government & public sector
Be the source AI cites for your own services.
“Residents ask ChatGPT how to renew a permit, and it quotes a fee we changed two years ago.”
Residents now ask ChatGPT, Gemini, Perplexity, Copilot and Google AI Mode how to renew a licence or whether a benefit applies to them. The engines assemble the answer from whatever they can read, and your own site is often the source they cite least. ELEVA audits your service pages, tracks those questions every day, reads what comes back, and hands your digital and communications teams a roadmap.
The problem
The engines answer for you, from sources you do not control
Five ways a resident gets the wrong answer about a public service today.
The answer quotes a fee you changed two years ago
Someone asks Copilot what a business licence renewal costs and gets the figure from a fee schedule PDF that is still online. The current fee is on a page the engine did not read, or read and could not quote.
A commercial “how to” site is cited ahead of your page
The engine builds its answer from a commercial guide, a local news story and a Reddit thread, because each of them states the steps in one place in plain words. Your page links to a portal and a PDF, so the engine cites the site that paraphrased you.
Your robots.txt keeps the engines out, so they read someone else
A rule added in 2023 blocks the AI crawlers from the whole domain. Residents keep asking anyway, and the answer gets assembled from whatever else is online about your services, without your side of it.
The page never says who is eligible, what it costs or when it closes
Eligibility is in a PDF and the fee is in a portal behind a login; the deadline went out in a press release. No single page puts the cost, the steps and the date together, so no engine can quote one.
The rule changes and the answer does not
An office closes or a deadline moves, and communications publishes the notice. For months afterwards the engines keep citing the old page and the news story about the old rule, because nobody is assigned to notice.
Why now
Residents already ask the engines. Other people’s research shows what they get back.
42%
of US adults use an AI chatbot to search for information
About half of US adults now use a chatbot at all, up from a third in 2024, and six in ten read the AI summary at the top of search results.
45%
of AI assistant answers to news questions had at least one significant issue
31% had serious sourcing problems and 20% major accuracy issues, including outdated information. Over 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity, 22 public service media organisations, 18 countries.
39%
of ChatGPT answers to voter questions pointed to the official state election site
Google AI Overviews did so in 56% of answers, until a format change in February 2026 dropped that to zero. The engines were accurate and still did not send people to the authoritative source.
60%+
of citation queries answered wrongly by eight AI search engines
1,600 queries. More than half of the answers from two of the engines cited fabricated or broken URLs, and most of the wrong answers were given confidently.
First, the templates and the files
An audit of the pages residents land on, and the files the engines read first.
A public site is a handful of templates repeated thousands of times: the service page, the fee schedule PDF, the portal, the news release, the location page. We audit one URL of each with 21 checks: whether the 20 AI crawlers we track are allowed in, whether the text is in the page or drawn in by scripts, whether the page states the cost, the steps, the documents and the deadline in words an engine can quote, and whether robots.txt, llms.txt and the sitemap agree with the pages.
You get a finding per template, with the fix written for the people who own it: what web unblocks, what the service owner rewrites, which PDFs get a page or a retirement date. Your teams ship it, and we recheck and score it after.
- Crawler access for 20 AI bots, checked per template
- robots.txt, llms.txt, sitemap and canonicals read together
- PDFs and portals that hold the facts the pages do not
Then, the questions
The questions residents ask about each service, put to every engine daily.
For each service we build the question set residents type: how to renew, what documents it needs, who is eligible, what it costs, when it closes, and where the office is. The platform runs them every day against ChatGPT, Gemini, Perplexity, Copilot and Google AI Mode, in the country residents ask from, and records the full answer and every source it cited.
From that you see which of your pages get cited and how often, and which other sites, news stories and threads the engines trust instead. When a commercial guide is cited for your permit process more often than your own page, it shows up as a row, with the URL.
- Cited pages and cited domains, split into yours and everyone else’s
- Tracked in the United States, the United Kingdom, Brazil and Greece today
Read by a person
An accuracy read of what the engines tell residents, claim by claim.
The platform keeps every answer. A person on our team reads them, engine by engine, and flags each claim as correct, outdated or wrong: the fee, the deadline, the documents, the office, the rule. Each flag is traced to the source the answer cited, so your team knows whether the problem is a stale PDF on your own domain, a third-party guide, or a page the engine never reached.
No software scores this. A reviewer sits with your service catalogue open and reads, and that is the part a public body rarely has anyone free to do.
- Every wrong or outdated claim, with the engine and the cited source
- Rechecked daily after your team ships the fix
Roadmap and briefings
A roadmap with owners, and a monthly briefing on what changed.
Findings become a roadmap with an owner per item. The digital team unblocks the crawlers and fixes the templates, the service owner rewrites the page to state the fee and the deadline, and communications retires the old PDF and publishes the notice where the engines will read it. The list is ranked by the questions residents are getting wrong most.
Every month you get a short briefing in plain language: which of your pages the engines started citing, which claims were corrected and which still stand, which sources gained or lost, and which new questions residents are asking. It is written for the people who report upward.
Where it applies
The public bodies this is written for
Any body whose residents, riders or customers ask an engine how a service works. For most the goal is to be the cited source. For two it is also to be recommended.
National and regional agencies
Tax, benefits, licensing, immigration and registries. These carry the most questions and the most PDFs, and a wrong answer costs the most.
Cities and municipalities
Permits, parking, waste, council tax and planning. The engines answer these from a commercial guide or a neighbourhood thread.
Public health services
Where to get vaccinated, what a clinic treats and who qualifies for a programme, which residents act on the same day they ask.
Transit authorities
Fares, passes, concessions, and what changed on the line this month. Riders ask the engine before they ask the app.
Public utilities
Tariffs, outages, connections and payment help. The fee the engine quotes is the one on the old tariff PDF.
Tourism boards and economic development
The one place a public body competes to be recommended, whether the question is where to relocate a business or what there is to do in your region.
The engagement
GEO Consulting & Trends, quoted per engagement
Your digital and communications teams execute. We bring the audit, the tracking, the accuracy read, the roadmap and the briefings.
For public bodies
GEO Consulting & Trends
ELEVA is a small agency. Each engagement is quoted after a free audit, and it can run as a fixed-scope project with a defined deliverable list, which is how public procurement usually buys, or as a retainer. A small municipality, or a single agency with no web team, that wants its pages rewritten and published for it is a Managed GEO engagement instead.
- A free AI visibility audit of one URL per template, with a before-and-after score
- The audit of the service templates and the files: robots.txt, llms.txt, sitemap, PDFs, portals
- The questions residents ask per service, run daily across every major AI engine
- An accuracy read of the recorded answers by a person, with every wrong claim and its cited source
- Cited pages and cited domains: your pages against the sites cited instead of them
- A roadmap with an owner per item for the digital, service and communications teams
- A monthly briefing on what changed, written for the people who report upward
The platform behind the work
What your teams log in to
- 21 checks on any URL: access for 20 AI crawlers, server-rendered text, content depth, and the files the engines read first.
GEO Audits
- The questions residents ask, run daily per engine and country, with the full answer and every cited source kept.
Prompt Tracking
- Saved views per service and per team, cut by engine, country and tag, and compared with the previous window.
Custom Reports
From the first call
What public sector teams ask us
Can we buy this as a fixed-scope project rather than a retainer?
Yes. GEO Consulting & Trends is quoted per engagement, and for a public body that usually means a project with a defined deliverable list: the audit of the templates and the files, a question set per service tracked for an agreed period, the accuracy read, the roadmap, and a set number of briefings. We are a small agency, so we will say plainly what we can take on and what we cannot. If your procurement needs a particular document format or a supplier registration, say so in the first message and we will work through it before we quote.
Can you make the engines stop giving residents the wrong fee?
Nobody can edit an engine. We can take away the reasons it gets the fee wrong, which is what the roadmap does. The page with the current fee has to be reachable by the crawlers, state the figure in plain text rather than in a PDF or behind a portal, and be the page your sitemap and llms.txt point at. The old fee schedule has to be retired or redirected, because it is the source the answer cited. The accuracy read finds each wrong claim and its source, and the tracking rechecks the question daily once your team ships the fix. Engines recrawl on their own schedule, so we will not promise a date.
Should we block AI crawlers in robots.txt?
That is your decision, and we do not give legal advice on it. The audit gives you the fact base: which of the 20 AI crawlers we track are allowed in, per template, and what each of the five engines is citing for your services today. Residents ask either way; with a block in place, the answer about your permit process comes from a commercial guide and a news story instead of your page. One common choice is to allow the crawlers on the service pages and keep them off the rest. Whatever you decide goes into the roadmap, with an owner.
How do you check whether an answer is accurate?
A person reads it. The platform records every answer to every tracked question, engine by engine, with the sources it cited, and a reviewer on our team goes through them with your service catalogue open. Each claim gets flagged as correct, outdated or wrong, and each flag is traced back to the source the answer cited. There is no automated accuracy score, and we do not think one would be honest. You get the list, and the tracking tells you when a fixed claim stops appearing.
We are a small municipality with no web team. Is this still for us?
Yes, as Managed GEO rather than consulting. The consulting engagement assumes there is a digital team and a communications team to execute the roadmap. Where there is not, we do the work ourselves: fix the crawler access and the templates, rewrite the service pages to state the fee, the steps, the documents and the deadline, publish them on your site, and track the questions daily. Managed GEO starts at $990 a month and is scoped after the free audit. It also fits a single agency inside a larger body that wants its own pages done.
See what AI is already saying about your brand.
A free audit shows you exactly where you stand across ChatGPT, Gemini, Perplexity, Copilot, Claude, Grok, and AI Overviews, and what it would take to close the gap.