40 Questions Politicians Should Ask Before Using AI for Speeches, Policy, and Constituent Communications
When you use AI to draft a speech or analyse policy, you risk outsourcing the deliberative work that voters elected you to do. The right questions protect your authenticity and keep your judgement in the driver's seat.
These are suggestions. Use the ones that fit your situation.
Questions to ask before using AI for speech writing
1Can I point to a specific experience or conviction in my own life that would support this argument, or did the AI generate it from pattern matching?
2Does this phrasing sound like how I actually speak to constituents at the doorstep, or does it sound like polished rhetoric from a broadcast news segment?
3If someone from my community read this speech and asked me to explain the reasoning, could I defend every claim without reference to the AI's sources?
4Have I checked whether the AI misrepresented a local fact, statistic, or the history of an issue in this community?
5Is there a position in this speech that contradicts something I have said publicly before, and if so, have I made a deliberate choice to change my view?
6Does this speech address the actual concern my constituents raised at last month's town hall, or does it address a concern the AI assumed based on demographic data?
7Could a journalist or opponent credibly argue that AI wrote this speech by finding identical phrases in other politicians' recent remarks?
8Have I removed or rewritten the sections where the AI made emotional appeals that do not match the actual stakes of the issue?
9If I deliver this speech, am I comfortable being asked in the next interview whether AI had a hand in writing it?
10Have I read this speech aloud and edited it until it flows the way I would naturally speak, rather than the way a language model writes?
Questions to ask before using AI for policy research and positions
11Did I form my own view of what the right policy outcome is before I asked the AI to research arguments in favour of it?
12Has the AI's analysis of this policy area shaped my position, or have I used it only to fill in gaps in research I already had a direction on?
13Can I name two or three people in my community whose lived experience with this issue informs my stance, or am I relying entirely on AI summaries of data?
14Does the AI's recommended policy position align with the values I campaigned on, or does it represent a compromise I did not consciously choose?
15Have I tested the AI's analysis against advice from local experts, advocates, and community leaders who actually work on this issue?
16Is the AI citing sources, or is it generating plausible-sounding policy reasoning that I cannot verify?
17If this policy position backfires, will I be able to explain to my constituents why I chose it, without saying an AI recommended it?
18Have I deliberately chosen to disagree with the AI's recommendation in any part of my position, or have I adopted its analysis wholesale?
19Does the policy analysis account for how this issue affects my specific district, or is it generic analysis that could apply to any constituency?
20Have I discussed this policy position with the party research team or my senior advisers before committing to it publicly?
Questions to ask before using AI for constituent communication at scale
21Would a constituent recognise this message as something I actually wrote to them, or does the personalisation by AI make it feel templated?
22Have I checked whether the AI personalisation is based on publicly available information about how this constituent voted, rather than something they told me directly?
23If this constituent responds to my AI-personalised message, will I or a real staff member read their reply and answer them thoughtfully, or will they get another AI response?
24Am I using AI to send messages to constituents I have never met at scale, and if so, am I being transparent that these are automated?
25Does the AI address this constituent by a name or nickname they actually use, or has it made an assumption that feels oddly familiar?
26Have I sent a test version of this AI-personalised message to a handful of real constituents to see whether they felt the relationship was authentic?
27Is the AI-personalised message actually addressing the individual constituent's question or concern, or is it sending a generic reply dressed up with their name?
28If the media asks whether I used AI to personalise constituent messages, will my answer strengthen or weaken voter trust?
29Have I instructed my staff that certain types of constituent queries (complaints, requests for help, urgent concerns) should be handled by a human, not an AI?
30Am I using AI personalisation because it genuinely improves communication with constituents, or because it allows me to respond to more messages without hiring more staff?
Questions to ask before using AI for media strategy and constituent feedback analysis
31When the AI identifies an emerging issue in constituent feedback, have I verified this against my own reading of local news and conversations with people in the community?
32Is the AI analysing what constituents actually care about, or is it pattern matching against content from other regions that happen to be similar in demographic profile?
33If I take the AI's advice on which issues to emphasise in my media strategy, am I still making space for the issues my most vulnerable constituents care about?
34Have I checked whether the AI's recommendation to shift focus away from a particular issue is because it detects low engagement, or because a few vocal critics are pushing back?
35Does the AI's sentiment analysis of social media accurately reflect the tone of actual conversations in my constituency, or is it picking up noise from outside the district?
36Am I using Quorum AI to monitor what other politicians are saying so I can copy their approach, or so I can understand legitimate alternative positions I should respond to?
37If the AI recommends that I avoid or minimise a particular policy position because it polls badly, have I considered that my role is to lead, not just to follow constituent opinion?
38Have I set clear boundaries with my staff that AI-driven media strategy should inform my thinking, not replace my political judgement about what message I should lead with?
39When Perplexity surfaces research showing strong constituent support for a position, have I read the source material itself, or am I trusting the AI's summary?
40Is the AI telling me that constituent concerns are shifting, or is it telling me that online engagement is shifting, and have I confirmed the difference with real conversations?
How to use these questions
Before you run an AI-generated speech or communication through final approval, read it aloud as if you were hearing it for the first time. If it sounds like someone else, it is.
When an AI tool analyses constituent feedback or policy research, treat it as a starting point for your own thinking, not as a finished analysis. Your local knowledge is irreplaceable.
Ask yourself: can I defend this decision in the next election debate? If the answer depends on hiding that AI was involved, you have made a mistake.
Brief your communications team that any AI-generated content for constituents must pass a simple test: would this person believe I wrote this myself, or would they sense it came from a machine?
Set a rule with yourself that you personally draft the opening and closing lines of any major speech, and that you personally review policy positions before they go public. This keeps your judgment active and present.