Answer Engine Optimization

AI Narratives

Your voters are asking ChatGPT, Gemini, Claude, and Perplexity who to vote for. AI Narratives shows you exactly what the AI tells them about your candidate — and how to change the answer.

Get Your AI Narrative See a Real Example

Your next voter is asking an AI

Search used to hand voters ten blue links and let them make up their own minds. An AI assistant hands them a single, confident paragraph — and that paragraph is the impression. If the model confuses your candidate with someone else, leads with an old controversy, or leaves them off the ballot entirely, that is what your voter walks away believing. You never see it, and until now you couldn't do anything about it.

4
major assistants — ChatGPT, Gemini, Claude, Perplexity — now answer “who should I vote for?” directly
1
answer per question. No ten links to argue with — the model decides what voters hear
~1 day
how fast the quickest engines re-read a corrected page and change their answer about you

We ask the questions your voters ask

We probe all four major engines — both the free and flagship models, with live web search — using the real questions voters and donors type. Every answer is graded helps neutral hurts for your race, with the grader’s full reasoning. Then we hand you the exact pages to publish and the profiles to fix so the engines change their answer.

Identity & news

“Who is [candidate]?” · “What’s the latest news?” — does the AI know who they are, and what does it lead with?

The ballot

“Who’s running?” · “Who should I vote for?” — is your candidate even on the list the model gives?

The issues

“Where do they stand on the economy, healthcare, abortion, immigration, guns?” — is the position right, and does it land?

The persuadable voter

“I care about working families — who should I vote for?” — how the AI frames your candidate to the voters you need most.

A real AI Narrative

What we found for a first-time U.S. Senate candidate in a crowded primary. The name, state, and identifying details are removed — the grid and the findings are the actual report.

A question your voter asksChatGPTGeminiClaudePerplexity
Who is the candidate?neutralhelpshelpsneutral
Latest newshurtshurtsneutralneutral
Who is running?neutralneutralhurtsneutral
Who should I vote for?neutralneutralneutralhurts
Progressive voter: who to back?neutralneutralneutralneutral
Conservative voter: who to back?neutralhurtsneutralhurts
Good candidate?neutralhelpshelpsneutral
Economy & inflationhelpshelpshelpshelps
Healthcarehelpshelpshelpshelps
Abortionhurtshurtshelpsneutral
Immigrationneutralhelpsneutralhurts
Gunshurtshelpsneutralhurts
helps builds supportneutral no effecthurts costs you votes

The biggest problem: the AI didn’t know who he was

One engine confused him with three other public figures who share his name; another repeatedly named the wrong person as the race’s Democratic challenger and left him off the ballot entirely. For a candidate trying to break out of a crowded primary, being treated as a footnote — or the wrong person — was the most damaging finding. It was also the most fixable.

Where he won: the kitchen-table message

Economy and healthcare came back helps on all four engines — working families, manufacturing, lowering costs, rural hospitals. When the models engaged with his actual platform, the message landed cleanly everywhere.

Where he was exposed

“Latest news” hurt him on the two biggest consumer engines, both leading with a single controversy instead of his platform. Abortion and guns read inconsistently — helping on one engine, hurting on another — because no authoritative page stated his position for the models to cite.

Why it happened

Two neutral reference sources did almost all the heavy lifting for the engines, and his entries there were thin or empty. His campaign site was nearly all biography with no issue pages — so the models had nothing authoritative to cite for where he stood.

The fixes we handed over

  1. Fix the identity problem first. Expand the authoritative reference entries and the disambiguation line so the engines stop confusing him with same-named people or omitting him from the race.
  2. Publish “Where [candidate] stands on …” pages for the weak issues — abortion, immigration, guns — each backed by a statistic and a named source. The single biggest visibility lift.
  3. Add a real issues page to the campaign site with citations, so the fastest-refreshing engines have something current to pull within about a day.
  4. Complete the neutral third-party profiles the engines trust for candidate facts, so every model is reading the same verified record.
Client Zero

We don’t just sell it. We ran it on ourselves.

Before we offered AI Narratives to a single campaign, we pointed it at Civly. The verdict was humbling: the assistants kept confusing us with an unrelated co-parenting app, couldn’t find our pricing, and in one case told the user no such company existed. So we did exactly what we now do for clients.

Before“There is no known entity called ‘Civly’.” — a major engine, treating our name as a dictionary word.
AfterA structured, sourced answer that names the right Civly — the AI campaign-intelligence platform — and what it actually does.

We published a structured Wikidata record so the engines have an authoritative, machine-readable account of who we are, and completed our Crunchbase profile so the reference sources AI trusts for company facts point to the right company. It’s the same playbook we hand campaigns.

↗ Our Wikidata record ↗ Our Crunchbase profile

We believe in this enough to make ourselves client zero.

Find out what AI is telling your voters

We’ll run your candidate through every major assistant, grade every answer, and hand you the plan to change the ones that hurt.

Get Your AI Narrative

“If the AI can’t say who you are, your voters can’t either.”