AI is already answering "who should I vote for?"

Make sure the answer is accurate, sourced, and yours. Visigent for campaigns, advocacy and policy.

5AI assistants checked, up to: ChatGPT, Gemini, Perplexity, Claude, Grok

24hto your GEO Score and first three fixes

$5000a month to start.

0changes published without your approval

Visigent for politics

AI is already answering "who should I vote for?"

Voters ask ChatGPT, Gemini, Perplexity and Grok about candidates, measures and issues, and get one answer stated like fact. Today that answer leans on news coverage and whoever showed up first. We make sure it draws on your record, framed accurately, and we measure the change.

Call for pricing Get your free GEO score

Pricing on a call · Free index in 24 hours · Any party, one side per race

AI doesn't choose sides. It chooses sources.

Why now.

Whoever owns the answer owns the next generation of voters.

Voters

1 in 5

voters now consult AI chatbots for election news and voting guidance, Democrats and Republicans at similar rates. AAPC Foundation survey, September 2026

Reach

41%

of likely voters saw political information in an AI search summary in the past month. 17% asked a chatbot directly. Tavern Research for Caucus AI, n=4,314, August 2026

Gaps

88.9%

of ChatGPT answers about governor's races gave an incomplete list of candidates. Accurate is not the same as adequate; a missing name is a missing vote. States United Democracy Center, August 2026

Trust

1 in 2

people who get news from chatbots say they at least sometimes see news they think is inaccurate; 59% of those under 30. The answer is believed, and it's often wrong. Pew Research Center, October 2025

Three families of questions decide a race.

We map and track hundreds of prompts in each, on every major engine, every week.

Recommendation

Who should I vote for?

“Who is the best candidate for affordability in North Carolina?”
“Which group protects hunters' rights?”
“Should I vote yes on Proposition 12?”

Position

Where do they stand?

“What does [candidate] believe about taxes?”
“[Organization]'s stance on wolf delisting”
“What would [measure] actually do?”

Head-to-head

Compared to whom?

“[Candidate A] vs [Candidate B] on public safety”
“Alternative to the incumbent on housing”
“Who really funds conservation?”

How an AI engine decides what's true about a candidate.

Two pillars of information, and a hierarchy of who it believes.

Pillar one · what you publish

Your website is the knowledge base. Most campaign sites are one or two pages built to raise money; engines need positions, record, people, places and numbers, each on its own page.
Structure matters more than prose. Engines want simple structure, lists, high detail and no hyperbole. "Veteran" and "voted against the tax increase" are facts; "fighter for families" is noise they discard.
Our system flags the loose descriptors engines can't use and replaces them with specifics before anything publishes.

Pillar two · what others say

A claim that exists only on your site gets omitted or hedged: "insufficient supporting information." Engines look for corroboration before they repeat anything.
They weigh sources in tiers: government and reputable research sites first; national earned media (wire services, major papers, broadcast) second; blogs, mid-tier outlets and community-curated platforms third; social posts and video below that.
Knowledge graphs (Wikidata and the data vendors behind them) are how engines decide you exist as an entity at all.

Two voters ask two different questions.

Both have to be answered, and they live on different pages.

Already knows you

"Should I vote for [name]?" "Where does [name] stand on housing?"

This voter got a ballot, saw a sign, heard the name. The answer has to be your record, framed accurately, with the proof attached. Won on your own site plus corroboration.

Still deciding

"Which candidate for governor aligns with my views on the border?"

This voter doesn't know your name yet. The answer is decided by who seeded the best-structured, best-corroborated answer to that question first. Won by answering thousands of questions before they're asked.

Defense and offense. Same engine, two jobs.

Most programs start on defense: fix what AI already gets wrong about you. Offense is where the lead is built: claim the questions nobody has answered before the other side does. We run both from one baseline and report them separately.

Defense · hold what's yours

Find and fix what AI gets wrong about you: a misstated position, a wrong date, a stale season or rule, a confusion with someone who shares your name
Pre-empt the misframes engines repeat, with a sourced fact-check that exists before the attack does
Hold your name: the knowledge panel, the prime image slots, page one for your own name
Watch for new wrong claims entering answers and respond while they're still rare, not after they harden into the default

Offense · take the open ground

Claim the questions nobody on your side has answered, written as specific, sourced statements, the form engines prefer to cite
Own the head-to-head: when a voter asks how you compare, the comparison comes from your record, not your opponent's
Win the recruiting and recommendation questions that grow a base, a membership or a vote
Go on offense on new questions before they're contested; the first credible source becomes the default, and defaults stick

Who it's for.

Candidates

Candidates and campaigns

Name search, knowledge panel, image results, and the answer AI gives when a voter asks where you stand.

For campaigns

Advocacy

Advocacy groups and coalitions

The policy questions you exist to answer, and the recruiting questions that grow your membership.

For advocacy

Measures

Ballot measures and initiatives

A fixed election date, a question voters ask AI by number, and an answer somebody will own.

For measures

Policy

Policy, public affairs and government relations

Think tanks, trade associations and corporate public affairs: owning the issue questions before the bill moves.

For policy

Pricing for political programs

Scoped by race, term and ground to cover. Priced on a call.

A primary, a ballot measure and a twelve-month advocacy program are different jobs. Thirty minutes with us and you leave with a scope, a number and the first three things we'd fix.

Call for pricing

How it works in politics.

Commercial GEO gets a business recommended. Political GEO makes an identity unambiguous and an answer owned before the other side writes it. Five moves, re-scanned every month. Nothing publishes without your sign-off.

Lock the baseline

Scan every engine across the three question families. Map each answer to whoever owns it today: you, the other side, a newspaper, Wikipedia, nobody. Week-one numbers you can re-run yourself.

Build the entity

A Wikidata item, Person or Organization schema, and one canonical photo free-licensed on Commons. The identity engines reach for before they reach for anything else. This is what splits you from the actor with your name and what prompts the knowledge panel.

Claim the open ground

Build the knowledge base, then answer the questions: hundreds to thousands of them, mapped in tiers from the platform down, each with a page a voter would recognize. Every fact held in Atlas with its source, so it says the same thing everywhere.

Become the source

Earned media through PR partners who have the relationships; placements, credentialed community participation and video where they help. Engines rank by the authority of the page hosting the fact, not by your site. Corroboration is what makes the answer cited, not just correct.

Hold it and prove it

Re-scan all engines monthly: share of voice, sentiment, answers owned, prime image slots, panel status. Defend ground as engines update. The scoreboard is yours; re-run it any month and check our work.

Who does the work.

People make the placements and the calls. Agents do the volume.

The team

Experienced Technical and Public Relations specialist

They build the pages, secure the placements, work with Wikipedia and community platforms, and pitch journalists. For high-value earned media we partner with PR firms that have the relationships; programmatic pitching alone converts at a fraction of a percent, and we say so.

The agents

Hundreds of AI Agents, one review

A research agent that gathers everything public about the candidate into one record. A question agent that maps voter questions in tiers, from the platform down. A writer. A page builder that matches your site's look and code. A checker that audits every page and placement against your messaging before anything goes live.

What it looks like.

One result and one baseline, names withheld during the cycle. Programs run across governor, Senate and congressional races.

National advocacy group · June 2026 scan1 of 5On the five questions that decide its issue, the group owned one answer. Opposing organizations owned three. Invisible on four of five engines; 13th of 15 groups on AI authority.Pre-program baselineWhat we proposed Gubernatorial candidate · Six monthsHundredsof Al answers the candidate now surfaces in. A corporate lead with no public presence and invisible to the engines at the start. Six months later: named across the engines on the questions that matter, and the party's nominee.Program result · name withheld during the raceHow it works for campaigns

The rules we work by.

Political work has a lower tolerance for shortcuts than retail. These aren't aspirations; they're terms.

Facts

Every claim traces to a source you can show.

Atlas holds the record — positions, votes, filings, numbers — with the document behind each one. Nothing goes live that you can't stand behind under oath.

Attribution

Your content is yours. Third-party sources stay third-party.

We never seed fake reviews, sock-puppet accounts, invented endorsements or "independent" sites you secretly control. AI engines detect it, and so does the press.

Conflicts

One side of a race. One side of a measure.

We don't work both sides of the same contest, and we say so in writing.

Compliance

Disclaimers, filings and coordination rules are yours.

We flag where paid-for language belongs. Your counsel decides. We don't touch fundraising data.

Tone

Correct, not combative.

The job is to make the answer accurate when someone asks about you. Attack content is a different business; it's not ours.

Community

Credentialed answers only on Reddit and the like.

Engines weight community-curated platforms because the community polices promotion. Accounts that lead with promotion get banned and the engines have already cut their weighting once. We participate as who we are, answering the question.

Scope

We don't write speeches.

We build the information layer: the record, the positions, the proof, structured so engines can find it. What the candidate says on a stage is the campaign's.

Any party

We work across the aisle.

The method is the same whoever asks. What we won't do is let a client's facts contaminate another's program.

Questions before you call.

Can you make ChatGPT say my candidate is the best?

No, and anyone who says they can is selling something. What we can do is make sure the engines have your actual positions, record and proof to draw on, in a form they read and cite, and measure whether the answer changes. In our experience it does, because today most answers lean on news coverage and your opponent's material.

Is this legal?

Publishing accurate, sourced information about a candidate, measure or policy position is ordinary campaign communication. Where disclaimers or filings apply, we flag it and your counsel decides. We don't do anything that depends on hiding who's speaking.

How fast does it move?

Foundation work lands in weeks: entity, canonical photo, structured positions. First movement in AI answers usually shows inside 45 to 90 days. A knowledge panel is Google's call and typically takes months. We report the trend monthly against a week-one baseline, so you see what moved and what didn't.

What is Atlas in a political context?

Atlas is the model of what your campaign or organization knows: positions, record, people, places, numbers, documents, each with a source. AI engines reward consistency and corroboration. Atlas is how one verified fact shows up the same way on every page, in every schema object, and in every third-party mention we earn.

Do you work with both parties?

Yes. We don't work both sides of the same race or measure.

What does it cost?

Political programs are priced on a call, because a primary, a ballot measure and a twelve-month advocacy program are different jobs. Most managed programs run $5,000 to $10,000 a month; larger scopes with paid media run higher. Thirty minutes gets you a scope and a number.

When should a campaign start?

Earlier than feels necessary. Engines build confidence in an entity over months, and corroboration takes time to earn. For a November race, March is comfortable; the summer is workable; the final six weeks is defense only.

We're a down-ballot race. Is this for us?

It's where the impact is largest. Almost nobody at the state-legislative or county level is competing for these answers, so a modest program surfaces a candidate fast. Our programs run across governor, Senate and congressional races, and we take smaller ones when the fit is right.

Reporters: barbee@visigent.com. No sales call required.

Start with the scan.

Enter your campaign or organization's website. In 24 hours: how each engine describes you, which questions you own, which you've lost, and the first three fixes.

Call for pricing Get your free GEO score

Call for pricing Free index