For policy & public affairs
The staffer asked ChatGPT first. Your position was not in the answer.
“We need to know how the engines frame the issue, in each country, before the hearing, not after.”
A legislative aide, a reporter, an analyst, or a voter asks an AI engine for the arguments on a rule, or for your organisation’s position on it. The answer picks a frame and names the sources it treats as authoritative. Where it describes your position, it sometimes does so in an opponent’s words. Your white papers and testimony sit in PDFs the engines never cite. ELEVA tracks what the engines say about your issues and your organisation in each market and for each persona, which sources shape the answer, and what to publish so it carries your evidence.
The gap
What you publish, and what the engines cite
Your evidence is written for a committee and a filing system. The engines read something else.
What your organisation publishes | What the engines cite |
|---|---|
| A 40-page white paper as a PDF, with the finding on page 23. | The Wikipedia article on the rule, and the sources its footnotes point at. |
| Written testimony, filed with the committee and posted as a scan. | A news explainer, and an opposing group’s two-page summary of the arguments. |
| A position statement in the press release archive, dated and buried. | A rival think tank’s brief with the claim in the first paragraph. |
| A media monitoring report that counts coverage of the issue. | Nothing that measures the answer itself. Nobody on the team is reading it. |
The problem
The issue gets framed before you get to the room
Every day, ChatGPT, Gemini, Perplexity, Copilot, and Google AI Mode explain your issues to the people who decide them. Five ways that goes wrong for a policy team.
The answer picks the frame
Ask an engine for the arguments for and against a rule and it decides which ones count, in what order, and who gets the last word. That frame reaches the staffer before your briefing does.
Your position, in an opponent’s words
Someone asks what your organisation thinks. The answer quotes the explainer a rival group wrote about you, because that was the clearest page on the topic, and your own statement, if it appears at all, comes second.
The evidence is in a PDF
Your white paper is a PDF. So is the economic study, and so is the testimony, with the finding on page 23. The engines cite whichever HTML page puts that finding in one paragraph, and it is not yours.
A different answer in every market
The UK answer cites the regulator and the BBC, the Brazilian one a newspaper and a Reddit thread, and the US one Wikipedia and a trade title. A global monitoring feed averages all of that into nothing.
Nobody reads the answer
Media monitoring counts coverage and social listening counts posts. Neither tells you that Gemini has called your position “industry opposition” for three months, or that a source it trusts changed its wording last week.
Why now
The people who shape your issue are already asking the engines
Third-party figures, linked to their sources.
10%
of people worldwide use an AI chatbot for news every week, up from 7% a year earlier
Across 48 markets. Trust in those answers sits at 20%, and 42% of users say they often click through to the original source. The rest take the answer as given.
82%
of journalists use at least one AI tool, and 47% use ChatGPT
897 journalists surveyed between January and March 2026, up from 77% the year before. The reporter covering your issue has likely asked an engine about it before calling you.
45%
of AI answers about the news had at least one significant issue
Sourcing was the biggest problem, in 31% of answers, across 3,000 responses in 18 countries and 14 languages. On a contested issue, getting the source wrong means getting the frame wrong.
55%
of ChatGPT answers cited Wikipedia in July 2025, under 20% by October
Across 230,000 prompts. Wikipedia is the reference the engines lean on for a contested issue, and how much they lean on it moves without notice.
What we do about it
We read the answers and work the sources
We do not promise to make an engine take a side, and we do not spin. We show you how each engine frames your issue today, which sources it trusts, and what to publish so your evidence is one of them.
The framing
How each engine frames your issue, per market and per persona
We build the question set for each issue the way the people around it ask it. A journalist wants the arguments on both sides; a legislative aide wants to know who supports the bill; an analyst asks what the evidence says; a voter asks what the rule means for them. Each question runs once a day on every major AI engine, in each market you name.
We keep every answer in full and read it the way a staffer would: what frame it opens with, which organisations it names, whose words it uses for your position, and which sources it cites for each claim.
- Prompts per persona and per country: today the platform runs the United States, the United Kingdom, Brazil, and Greece
- Mention rate and position for your organisation and every other body the answers name
The sources
Which sources the engines treat as authoritative, and where yours are
The platform logs every citation in every answer: the page, the domain, and where in the answer it sits. The tables of cited domains and cited pages show what the engines lean on for your issue, ranked, with each row tagged as yours, another named organisation’s, or a third party’s.
The PDF problem shows up here. The rival brief, the opposing group’s explainer, the Wikipedia article, and the news piece all rank in the table. Your testimony and your white paper do not, and the table shows what the cited pages have that your documents lack.
- Cited domains and cited pages, with share and change per window
- Reddit, YouTube, and LinkedIn pages ranked, because the engines cite the thread about your issue too
The briefing
A monthly read on framing shifts, sources moving, and new questions
Each month your strategist reads the window against the one before it, market by market. The movers tables flag the sources gaining or losing share in the answers, the organisations named more or less often, and the questions that are new.
The briefing says what changed and why: a regulator’s consultation page is now cited in three engines, a subreddit is trusted in the US but not the UK, the opposing group’s explainer lost share to a news piece, people are asking a question about your issue that nobody tracks. Each signal names the team that should act on it.
- Movers per window: sources, organisations, and prompts that gained or lost
- Saved report views per market, engine, persona, and issue, shared by link
The roadmap
What to publish, and which sources to brief or earn
The roadmap goes to your policy, comms, and web teams with an owner on every line. Most of it is publishing: a position page in HTML that states your view and the evidence in the first paragraph, with the PDF linked under it rather than instead of it. One page per issue and per market, in that market’s language, written so it can be quoted.
The rest is source work. The briefing names the reporters, explainers, and reference pages the engines lean on for your issue, so your team can brief them with the evidence or offer the data they are missing. Your teams execute, and the daily runs show which answers moved.
- A GEO audit of your site: crawler access for 20 AI bots, rendering, structure, and whether your positions exist as pages
- Sources to brief, ranked by how often the answers cite them rather than by their reach
The engagement
Strategy for the team that owns the position
You have the policy staff, the comms team, and the relationships. We bring what the engines answer today and the sources behind it, with a roadmap your teams can execute.
For policy, public affairs, government relations, and advocacy teams
GEO Consulting & Trends
Quoted per engagement, as a retainer or a project, after a free audit of what the engines say about your issues and your organisation today. A small advocacy organisation with no web team can run Managed GEO instead, where our team builds and publishes the position pages.
- An audit of how every major AI engine frames your issues and describes your organisation, per market
- A question set per issue, persona, and country, run daily with every answer and citation kept
- Cited domains and cited pages, ranked and tagged as yours, another organisation’s, or a third party’s
- Movers that flag a source, an organisation, or a question gaining or losing, per window
- A monthly briefing on framing shifts and sources moving, with actions and owners
- A roadmap for policy, comms, and web: what to publish as a page, which sources to brief
- Saved report views per issue and market, and a dedicated strategist
The platform behind the work
What your strategist is looking at
- The questions a staffer, a reporter, or a voter asks about your issue, run daily on every engine, with every answer and citation kept.
Prompt Tracking
- The same issue read per persona: where the journalist’s answer names you and the legislative aide’s does not.
Visibility by Persona
- Saved views per issue, market, engine, and persona, compared with the previous window and shared by link.
Custom Reports
Questions policy leads ask us
Before you take it to the board
Will you make the engines take our side?
No, and nobody can. An engine assembles its answer from the sources it trusts, and it decides which those are. We show you how each engine frames your issue today, which sources shape that frame, and where your evidence is missing. The work after that is publishing your positions and evidence in a form the engines can cite, and briefing the sources they already lean on. When an answer changes, it is because a better source exists, and the daily runs show you which one.
Why is our testimony never cited?
Usually because it is a PDF. Engines that search the web prefer an HTML page that states the point in its first paragraph, with a heading structure they can parse and a date they can read. A scanned filing or a 40-page report with the finding on page 23 rarely gets quoted, even when it is the best evidence available. The audit checks whether your site can be crawled and read at all, and the roadmap turns each position into a page the engines can cite, with the PDF linked underneath.
Do you track what the engines say about the other side?
Yes. We track the organisations the answers name beside yours, whether allies, opponents, or a rival think tank, as a named set, so you see their mention rate, how early they appear, and which of their pages get cited, next to yours. Their explainer ranking above your position statement is usually one of the first things the sources table shows. We report what the engines say about them; we do not write anything about them.
We are a small advocacy group with no web team. Is this for us?
Probably as Managed GEO rather than consulting. The tracking and the monthly read are the same. The difference is that our team builds and publishes the position pages, fixes the site, and writes the explainers, with you approving every draft before it goes live. It starts at US$990 a month, scoped after the free audit. If you have a comms team that can publish, the consulting engagement hands them the roadmap instead.
How long until the answers reflect our position?
We will not give you a date, because the parts move at different speeds. A position page can be crawled within weeks, and an engine that searches the web can cite it soon after. Engines that rely more on what they were trained on take longer. The slowest part is the sources you do not own. A news explainer, a reference article, or a regulator’s page updates on its own schedule. From the first month you watch each question, engine by engine and market by market, beside the list of what your teams published.
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.