Fwd: Why Your Chatbot Is Lying About Public Affairs

Begin forwarded message:

From: Neptune Ops <info@neptuneops.com>
Date: October 12, 2026 at 8:05:03 AM EDT
To: mark@rauterkus.com
Subject: Why Your Chatbot Is Lying About Public Affairs
Reply-To: info@neptuneops.com



🔱 Three blind spots in chatbots, and how to spot the vendors selling them ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏  ͏ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­ ­  

View in browser October 12, 2026 4 Minute Read 🔱 Sign up for a free demo today!   See how the latest AI and digital advocacy tools can transform your work. THIS WEEK'S EDITION Why Your Chatbot Is Lying About Public Affairs  Ask ChatGPT, Claude, or Gemini who moves a senator on energy policy, and you’ll get a confident, well-formatted answer in seconds. It will read like a polished briefing memo. It will also be flat wrong. That’s not because the chatbot is trying to mislead you. Chatbots were built to serve hundreds of millions of people asking about everything from recipes to tax law. They were not built to map influence inside a capitol. In today’s short read, we cover the three reasons chatbots fall short in public affairs, what purpose-built tools do differently, and how to tell a real public affairs AI tool from a chatbot in a nice suit. The Three Chatbot Blind Spots   🔱 It won’t go there. Chatbots are designed to steer clear of sensitive political territory. Ask who an elected official’s closest influencers are, meaning the donors, former staffers, family, friends, and local leaders whose calls actually get returned. You’ll get a polite deflection or a generic list of committee chairs.   That’s intentional. The companies behind these models would rather be cautious than controversial, so their guardrails treat political influence questions as a minefield. A chatbot is like the official tour guide at the Capitol. It can tell you the history of every statue in the rotunda, but it won’t tell you who’s meeting in the back room.   🔱 It only watches the highlight reel. A legislator can have thousands of floor speeches, votes, press releases, newsletters, and social posts on a single issue. A chatbot sees only two things: what it absorbed during training, which stops at a fixed date, and whatever its web search pulls back in a few seconds, usually a handful of top results. No coach prepares for a playoff opponent by watching SportsCenter. They watch the full game film, because that’s where the tendencies show up. Public affairs works the same way. The insight lives in the pattern: how a member talks generally, how they talk about a specific issue, what they've actually done on a topic.   🔱 It gives a different answer every time. Chatbots are probabilistic, not deterministic. In plain English, they build each answer by predicting the most likely next word. That means the same question asked twice can produce two different responses.   That’s a feature when you’re brainstorming ad copy or slogans. It’s a liability when you’re telling a client whether a swing vote is gettable. If a bank teller gave you two different balances for the same account, you’d switch banks. Your intelligence tools deserve the same standard. What Purpose-Built AI Tools Do Differently   Purpose-built public affairs AI is designed to solve all three problems: It synthesizes the full record rather than a sample, pulling in any statement, vote, and relationship on file. It’s built to explore the sensitive political questions practitioners ask every day, because that’s the entire point of the tool. Its answers are grounded in a defined body of verified data, so it becomes nearly deterministic. Ask the same question twice and you get the same answer. The serious tools get there by combining two things.   The first is open-weight models: AI models an organization can download and run on its own terms. Think of it as owning a car instead of hailing a rideshare. The rideshare driver decides which roads are off-limits. When you own the car, you choose the route.   The second is retrieval-augmented generation, or RAG, which turns every question into an open-book exam. Instead of answering from memory, the AI first looks things up in a curated library of verified information. You don’t need to know how the engine works to see why that matters.   Beware The Chatbot In A Trench Coat   As demand for public affairs AI grows, so does a crop of vendors who put a new logo on a general chatbot and call it a platform. Under the hood, it’s the same product you can buy for $20 a month. It has the same political guardrails, the same highlight-reel problem, and the same shifting answers. The only thing that’s different is the invoice. What This Means For Public Affairs   Your chatbot isn’t useless. It’s a strong writing assistant, a fast place to start research, and a great way to get past a blank page. But it is not an intelligence tool. Treating it like one means walking into meetings with a fraction of the picture and full confidence, which is a dangerous combination in a business built on credibility. The organizations that pull ahead over the next few years won’t be the ones with the most AI subscriptions. They’ll be the ones that know which tool to trust for which job, and that can tell the difference before the contract gets signed. Three Ways To Operationalize This Email   Ask these three questions of any vendor to separate the real thing from snake oil:   🔱 Who are your AI engineers? A company that builds AI employs people (almost always in the San Francisco Bay Area) who build AI. If the answer is a long pause or “we partner with a leading AI provider,” you’re likely buying a chatbot subscription with a markup.   🔱 What’s your model architecture? Ask how they navigate the political guardrails built into commercial models. A real builder will walk you through their AI architecture with confidence. A reseller will change the subject, much like the chatbot it’s reselling.   🔱 What have you actually shipped? Ask to see software running in production, used by real clients on real campaigns. A slick slide deck is not a product. 🔱 If you're enjoying this content, please consider forwarding this email to a colleague or friend.    🔱 If you're not already a subscriber, please sign up here to stay up to date on the latest developments in political technology.   🔱 If you have suggestions for future newsletters or any other feedback, we'd love to hear from you. Neptune Ops, 945 Market Street, Ste 501, San Francisco, CA 94103, United States