AI Answer Intelligence for Public Affairs
Give AI clear, attributable information about the issues you represent.
AI systems now explain policies and public issues to your stakeholders. AITWIRE tests how they answer the prompts that matter, compares those answers with your approved positions and evidence, publishes clear, attributable information in formats AI systems can access, and measures changes over time. Accurate, transparent representation — never automated persuasion.
Checked independently with controlled prompts across eleven leading public AI systems and answer surfaces.
“What would the proposed amendment change for small operators?”
Re-measured after calibration against the same fixed prompt set — reported as Measured improvement, Not yet conclusive or one of the other canonical states.
Illustrative example output — not live data.
The questions have moved
What AI is telling your stakeholders.
Journalists, staffers, members and the public increasingly put the first question to an AI system rather than to a website. The answer they get is assembled from whatever the system retrieved — and the gaps that matter are rarely gaps in volume.
“What would this regulation actually do?”
Framing gapAn AI system summarises a bill from whichever sources it retrieved. If your reading of the provision is not available in a form it can use, the summary is assembled from everyone else's.
“Who supports this, and who opposes it?”
Attribution gapPositions get attached to organizations from press coverage and secondary commentary. Coalitions are inferred. An unattributed position is one that cannot be checked — by the stakeholder or by you.
“What evidence is there either way?”
Currency gapSuperseded studies and withdrawn figures keep circulating long after the record moves on. An answer built from a 2023 impact estimate is confidently wrong in a way that reads as authoritative.
AITWIRE does not control which sources an AI system uses. It makes your approved record available in formats those systems can access, and measures what they do with it.
Your approved record
The Issue Canon.
A structured, machine-first record of what you have actually said about an issue — approved by you, sourced throughout, and published in formats AI systems can access. It sits alongside the website you already operate; it does not replace it.
Nothing enters the canon without a person approving it, and every entry keeps its source and its effective date. That is what makes a position checkable months later, by a stakeholder or by you.
How the companion is publishedIssue Canon — record types
- Positions
- Your stated position on each issue, in your words, with the body that adopted it.
- Key claims
- The specific assertions supporting each position — separable, so each can be measured.
- Evidence
- The study, dataset, filing or testimony behind each claim, linked and dated.
- Definitions
- Contested terms defined explicitly, so an answer cannot quietly substitute a different meaning.
- Jurisdictions
- Where a position applies. A rule that holds in one jurisdiction is wrong in another.
- Counterargument responses
- Your response to the strongest opposing arguments, attributed and sourced.
- Effective dates
- When each position took effect and what it superseded — currency, made explicit.
Measure. Calibrate. Prove.
Measured like everything else on the platform.
Same loop, same instrument, same evidence threshold. A fixed verification set of prompts is re-run after every calibration, and the movement is reported with its confidence — including when there is not enough to conclude anything.
Position inclusion
Does the answer include your position at all, when the question is squarely about your issue?
Factual accuracy
Do the claims in the answer match your approved evidence — and where they diverge, how?
Framing
Whose characterisation of the issue is the answer using, and are qualifications preserved?
Source citations
What is cited, and does anything trace to your published, attributable record?
Stakeholder associations
Which organizations the answer associates with which positions — including yours.
Currency
Whether the answer reflects the current record or a superseded version of it.
Every calibration reports one of these
- Measured improvement
- No material change
- Adverse movement
- Not yet conclusive
- Insufficient evidence
Every re-measurement resolves to one of these five — “Not yet conclusive” included. It is a result we publish, not one we round up. See the methodology for how significance is decided.
Where the line is
The ethical boundary.
Advocacy work attracts requests this platform will not serve. The boundary is a product decision, so it is worth stating plainly before you ask.
What AITWIRE publishes
Attributable sources
Every published claim carries its source. A reader — or an AI system — can follow it back.
Clear sponsor identification
The organization behind a position is named in the record, not inferred from the domain.
Evidence separated from opinion
A finding and a view about that finding are different record types and stay that way.
Qualifications preserved
Scope conditions, confidence and limitations travel with the claim rather than being trimmed off.
What it will not do
No fabricated consensus
AITWIRE will not present a position as broadly held, or synthesise agreement that the record does not show.
No hidden personalization
One approved record, published openly. Never a different version of your position served to a different reader or system.
No persuasion optimization
The instrument measures how an issue is explained. It does not test message variants for what moves opinion furthest.
No generated advocacy prose
AITWIRE proposes facts and flags gaps. Your words stay yours, and nothing is published without your approval.
Start with the baseline
What is AI telling your stakeholders about your issue today?
Establish the baseline, find the gaps in framing, currency and attribution, approve the record and measure what changes.
Measure. Calibrate. Prove.