Précis AI Wants Policy Wonks

Précis AI Wants Policy Wonks

Opens Beta for Willard, the Policy-Native AI 

WASHINGTON, D.C., (Sept. 17, 2026) — To help address AI accuracy in policy work, Précis AI today opened a beta for Willard, a federal advocacy AI that provides direct, sourced answers drawn from verifiable public records.  

As Washington debates how to rein in powerful AIs, another risk is loose in the policy process: AIs like Claude and ChatGPT do policy work and produce fluent, confident, footnoted answers that are sometimes simply wrong.  

But AI is already reshaping policy work in Congress. POLITICO reported last month that the House Office of Legislative Counsel is being swamped by error-riddled AI-generated legislation. In that reporting, the former House legislative counsel warned that using error-prone AIs for statutory language creates “the highest risk at the apex of failure.” 

And the numbers back that up. Stanford’s 2026 AI Index found hallucination rates ranging from 22% to 94% across 26 leading models on a new accuracy benchmark — with many of the most capable models near the top of that range. Suprmind’s August 2026 hallucination rate tracking shows the newest flagships from OpenAI, Anthropic and Google hallucinate more than 60% of the time they don’t know the answer to a query, while OpenAI’s latest model has inaccuracies more than 90% of the time.  

“Speculation that AI will destroy humanity rules the airwaves, but the pressing problem is a confident, cited AI generation that’s just false,” said Précis AI CEO David Fuscus. “Willard is built to dramatically increase accuracy by directly accessing warehouses of government data.”  

General-purpose AIs generate mainly from training data and web sources; they aren’t built to ground every answer in official government documents. Willard is grounded in the world’s largest database of AI-accessible federal documents and engineered to retrieve, cite, and ground every answer in these public records.  

Beta testers gain early access, with a direct line to the product team to further customize Willard for policy. “Getting in early isn’t about access. It’s about shaping how the AI works,” continued Fuscus.  

Qualified professionals can sign up at precisai.com/Willard.  

About Précis AI 

Précis AI builds industry-specific AI tools. Précis Public Relations has users in 14 countries. Willard is a policy-focused AI designed to help, analyze and act on federal information for more accurate work. Details at PrecisAI.com. 

Media Contact 

Amelia McKee 

(202) 289-4001 

amckee@xenophonstrategies.com 

AI is transforming public relations, but it can also quietly amplify social biases. When an AI assistant only suggests male names for CEOs or filters out resumes based on gender, it isn’t just a technical glitch. These moments reveal deeper flaws built into today’s systems. Understanding these flaws is the first step to protecting your brand, clients and the audiences you serve. 

Data Bias: When History Repeats Itself in Code 

Data bias occurs when the information used to train AI is incomplete, unbalanced or reflects the prejudices of the past. This can happen through: 

  • Selection Bias: If a training dataset mostly features one demographic, then the model will struggle with others.  
  • Historical Bias: AI hiring tools have awarded jobs to men over women, mirroring an industry’s gender history and reinforcing outdated prejudices. 
  • Measurement Bias: When training focuses on a narrow set of variables, the model can ignore other components and create misleading results. 

Model Design Bias: When Assumptions Shape Outcomes 

The very structure of an AI model, such as what it prioritizes and how it weighs information, can create bias: 

  • Algorithmic Bias: An algorithm that boosts content simply for engagement might end up spotlighting divisive or harmful viewpoints over more informative ones. Recommendation engines, for example, can unintentionally amplify polarizing content or misinformation. 
  • Stereotyping Bias: Machine translation tools may default to associating “nurse” with “she” and “engineer” with “he,” turning subtle patterns in data into broad, real-world stereotypes. 

Developer Assumption Bias: The Human Factor Masked as Math 

AI is only as fair as the people building it. Bias sneaks in when developer perspectives shape how systems are built: 

  • Subjective Decisions: Even with good intentions, programmers might set benchmarks or filters based on their own worldviews, missing real-world diversity.  

Real-World Impact—More Than Theoretical Errors 

Bias in AI has moved from theory to front-page news: 

  • Predictive Policing: AI criminal justice systems trained on crime data from over-policed communities end up directing more police resources to those same neighborhoods, creating a self-fulfilling cycle of profiling. 
  • Healthcare Diagnostics: Medical imaging systems underperform for patients with darker skin because past data overlooked them, leading to potential misdiagnoses or missed care. 
  • Advertising Algorithms: Facebook’s ad system once showed lower-wage job ads more often to women and minorities, while high-paying positions favored white males. 

What Can PR Pros Do? 

Once you understand where AI bias comes from, the job shifts from awareness to defense. AI bias isn’t just a vendor’s problem — the moment a biased output goes public under your brand’s name, it becomes your reputation problem. Here’s how to get ahead of it. 

Push for guardrails above your pay grade. The single most effective protection isn’t at your keyboard — it’s organizational. Bring the risk to whoever procures AI for your company. Make the case for the infrastructure that catches bias before it ever reaches your desk: a clear AI-use policy, user training built around real, risk-based scenarios, mitigation tools that stop known failures before they ship, and quality-control oversight that keeps AI use inside regulatory and brand standards. Individual vigilance matters, but it works best backed by a system designed to prevent failures. 

Vet your vendors like a reputational risk. Before you sign on with an AI tool, treat data transparency as a condition of doing business. Ask where the training data comes from, how recently it was updated, whether it represents the full range of demographics your brand reaches, and what the vendor does to test for and correct bias. If they can’t answer, treat that silence as a risk signal — you’re the one who will answer for the output, not them. 

Keep a human in the loop on everything that ships. Automation should draft materials, never approve them. Build in a review step, so a person always checks AI-generated content for unintended stereotypes, tone problems or blind spots before it reaches the public. 

Pressure-test communications materials against your real audiences. Don’t take generic benchmarks at face value. Run materials through diverse, real-world scenarios that reflect the actual people your brand serves, so you surface bias in your context before your audience does. 

Make oversight a team sport. Run cross-team reviews with people of different backgrounds and experiences. Diverse human oversight is your single best tool for catching what a machine may treat as normal. 

The hidden bias problem in AI won’t fix itself. Download the Invisible Influencers Report to learn how to spot, fix, and prevent bias in your PR work. 

Government affairs has always been a business of people. Successful advocacy depends on trusted relationships, political judgment, strategic timing and the ability to navigate complex issues and legislative environments. Whether engaging lawmakers, building coalitions or advising organizational leadership, government affairs professionals bring experience and insight that no technology can replicate. 

Although fewer bills are being passed by Congress, what has changed in recent years is the sheer volume of information and the number of policies being proposed in legislation. In 2025 alone, more than 1,200 AI-related bills were introduced across U.S. states and territories, and that number has already been surpassed in 2026.  

While government affairs teams continue to monitor a growing number of regulatory proposals, agency actions, court decisions, stakeholder positions and media narratives across multiple jurisdictions simultaneously, the pace of policymaking is changing. As a result, organizations increasingly expect faster and more sophisticated analysis from their advocacy teams. 

At the same time, artificial intelligence has become one of the most significant policy issues facing lawmakers. More than 3,500 federal lobbyists—approximately one-quarter of all federal lobbyists—worked on AI issues in 2025, up from roughly 11% of all federal lobbyists in 2023. Data center lobbying has also grown by 500% over the past three years, reflecting the rapidly expanding policy ecosystem surrounding AI infrastructure and deployment. 

This is where artificial intelligence is beginning to play an important role. 

Scaling Government Affairs Using AI 

The most effective way to think about AI is not as a replacement for government affairs professionals, but as a workforce multiplier. 

AI can help monitor legislative and regulatory activity, summarize lengthy policy documents, identify emerging trends, prepare briefing materials and organize information at a speed that would be difficult for any team to match manually. 

The value is not that AI makes decisions, but rather that AI gives professionals more time to focus on the work that matters most—strategy, relationship-building, stakeholder engagement and advising decision-makers. 

As policy environments become more complex, AI helps government affairs teams spend less time searching for information and more time understanding what that information means and how it can be impactful. This capability is becoming increasingly valuable as organizations navigate a policy landscape where AI itself has become one of the most actively debated issues across federal and state governments. 

Human Advocacy Remains Essential 

While AI can process information, it cannot replace the human elements that drive successful advocacy. 

Technology cannot build trust with a legislator, negotiate competing stakeholder interests, understand political nuance or make strategic judgments during rapidly changing policy debates. Those responsibilities remain firmly in the hands of experienced professionals. 

The importance of human judgment is underscored by the growing number of stakeholders competing to shape AI policy. In 2025, the U.S. Chamber of Commerce employed 91 federal lobbyists working on AI issues, followed by Microsoft (63), Meta (55), Intuit (51) and Amazon (48). Overall, the top 30 AI lobbying organizations, and 91 of the top 100, were corporations or corporate trade associations, highlighting the increasingly competitive environment in which government affairs professionals operate. 

Organizations that are seeing the greatest value from AI are those that use it to support human expertise, not replace it. The most successful teams are combining technology-driven efficiency with the judgment and relationships that have always been at the core of effective government relations. 

Looking Ahead at AI in Public Affairs 

As AI continues to evolve, purpose-built platforms such as Willard from Précis AI are helping government affairs teams better manage legislative, regulatory and stakeholder intelligence that shapes today’s policy environment. 

The future of government affairs is not human versus machine. It is human expertise enhanced by technology. 

Policy developments are increasing in both volume and complexity, and AI gives professionals the ability to work smarter, respond faster and focus on the work that creates the greatest impact. As the policy conversation around AI continues to expand, organizations are also investing heavily in advocacy.  

During the first nine months of 2025, seven of the largest technology, AI and social media companies spent a combined $50 million on federal lobbying—an average of nearly $400,000 for every day Congress was in session—illustrating both the strategic importance of AI policy and the need for government affairs teams to keep pace with an increasingly dynamic landscape. 

Downloadable Use Cases for AI in Gov’t Affairs 

For those interested in exploring the practical applications of AI in government relations, a downloadable PDF outlining 25 specific use cases for professionals is available online. 

The resource highlights how AI can support everything from legislative monitoring and stakeholder engagement to compliance, forecasting and team enablement, while also reinforcing the principle that AI is designed to augment, not replace, relationship-driven advocacy. 

Many organizations have a “brand czar,” a marketing chief appointed to ensure the company speaks with one voice. AI is making that person’s job harder every day because AI “slop” allows employees to write exponentially more off-brand content every day. But a good AI platform can also make the chief marketer’s job astronomically easier by giving the entire organization the same set of up-to-date messaging to work from that match the company’s brand precisely. 
 
From email signatures to sales decks, the brand czar’s job has always been to slap the typing fingers of anyone in the company who fails to follow the official brand guidelines of the organization. Whether it be the CEO—and especially if it is the CEO—or an intern, everyone must have the same message if the organization has any hope of connecting with customers over the long term. Now, a marketer can upload the exact brand documents they want their teams to use as source material, so that every resulting document they produce will rely on the latest, approved messaging. Moreover, the brand czar can change the message on a moment’s notice and disseminate it across the organization in seconds if a crisis or opportunity arises. 

 
The Chicken Salad Test 

Southwest Airlines’ late chief Herb Kelleher, who had a nose for marketing, famously dictated to his company that they should focus on being “low-fare,” using the “chicken salad test.” As Chip and Dan Heath noted in their famous 2007 marketing book, Made to Stick, Kelleher reportedly said that if offering chicken salad wouldn’t make the company low-fare, then they shouldn’t do it. Taking a page from Kelleher’s wisdom, if a company’s marketing messages don’t reflect the brand, it will undo all the efforts to cultivate an audience. For instance, it would irrevocably harm Southwest’s brand value if the CEO were to brand the company as low-priced, but the sales team was selling a luxury product. The disjointed message wouldn’t ring true with the customer. 

How to Require “Message Discipline” with AI 

In the marketing profession, consistent repetition of brand value as Kelleher preached is known as “message discipline.” 

One of the biggest advantages a well-tailored AI platform can provide communications and marketing heads in this journey is the ability to force message discipline. A solid tool does this by giving all users access to the same platform, with cloud-based storage that is secure and private to the outside world.   

But there are some complications with this solution.  

Remove File-Sharing Friction  

The biggest problem is that many companies allow employees to choose any AI platform they please, or “phantom AI” tools. This means there is not a common enterprise version that everyone is using consistently, which makes file sharing complicated. Even if the company has a cloud system for sharing common docs on an intranet, or in a third-party app like Microsoft SharePoint or Google Drive, users may still have to migrate documents back and forth between their preferred AI platforms like ChatGPT or Perplexity. This friction causes employees not to use tools that cost hundreds or thousands of dollars per month. What a waste! 

Don’t Let Company Info Be Used as Training Data 

Another problem is data security. Some of the big AI providers like Claude are building incredible capability that integrates across Microsoft, Google, Adobe or other big, third-party providers. But there’s a concern here for organizations that prefer to keep their business or client information private. A lot of users simply cannot afford to allow their proprietary information to be used as AI training data by the big LLMs. In many cases, due to contractual obligation, it would be illegal for information to be shared with applications that inherently create such an information security vulnerability.  

Reduce the Threat of AI-Enhanced “Slop” 

A third issue that inhibits a brand czar’s ability to force message discipline is that some employees will inevitably choose to use AI to create tons and tons of garbage materials—or “slop.” This is one of AI’s biggest threats to reputation and brand, and it is among the hardest to control.  One recent Stanford University study showed that the more powerful an AI model is, the more it is likely to hallucinate, or make up false and, thereby, off-brand content. 

One way an AI platform can help combat this kind of slop is through built-in fact-checking tools. Many common platforms don’t have these features that allow a user to check and doublecheck their work. So, the resulting products—which end up being produced in great volume by non-thinking employees—not only comes out factually inaccurate and wrong, but that material is often used and reused by both man and machine after it is spat out into the world.  

There is a great need for corporate AI programs to cut these terrible practices off at the pass, before they are used to disseminate bad information. 

Key Features of AI to Promote “Messages Discipline” 

While it is practically impossible to make every employee use good judgement, or to turn every employee into a brilliant writer overnight, AI does have the capability to help. 

Among the most important features to look for in an AI tool is a single, cloud-based interface with a common file sharing system that can be deployed across a user base or across teams within a user base. Even better, secure and reliable AI tools should have fact-checking features built in that forbid the production of documents with ethical or other concerns, and do not share data with training models. In the best case, the AI tool should use “multi-modal” fact checking, spreading the risk of inaccurate information across multiple AI models before delivering a result. A Google DeepMind study shows the performance of various leading models. 

Most importantly, if an AI tool can be well controlled by the brand czar, the company can be better assured that the source material employees rely on to create new documents is always the latest, up-to-date, approved messages the company leadership has reviewed, vetted and approved.

AI “foundation models” are large-scale AI systems trained on vast datasets that power generative search engines like ChatGPT, Gemini, Claude, Perplexity and Copilot are now reshaping how PR professionals create and distribute content. For communicators, this shift matters because these tools no longer return a list of blue links like in a standard Google search; instead, they synthesize direct answers and cite a handful of trusted sources. Winning visibility now means: 

  1. Being included in the plain-language AI summary. 
  1. Becoming one of the cited sources. 

The transition is happening faster than most PR professionals realize. In May of this year, Google announced a major change to its search functionality. Google and other major tech companies have replaced the familiar list of links with AI systems that generate instant, comprehensive answers, and this move from finding websites to getting direct responses changes everything about how PR teams create, distribute and measure content impact. The shift is no longer a forecast — it’s measurable: organic click-through rates (CTRs) drop 61% when an AI Overview appears for a query, and that number is rising. That means users are finding the answers without ever clicking through to your website. 

What are Foundation Models vs. Frontier Models?

Foundation models form the backbone of the AI transformation. These massive AI systems are algorithms, trained on vast datasets of information, to process results into plain language answers, and handle countless other tasks. They are considered “foundational” because they are broadly used for a wide set of tasks, require low levels of additional/user-side training and have few barriers to entry. Therefore, they can be easily adopted by everyone from large enterprises to individual users.  
 
Frontier models, on the other hand, push boundaries further by implementing more advanced capabilities, often using the foundational models as their base but pushing the application toward more specific use cases. The applications built on these models — from Google’s Gemini to Anthropic’s Claude — represent a new and emerging category of tools that understand, synthesize and create rather than simply retrieve information.  
 
Within the space of a few years, foundational models have become table stakes for communicators, while frontier models represent a cutting-edge future that promises to provide PR pros and marketers new capabilities that increase efficiency and enhance existing talents and skills. 

Which AI Search Engines Matter for PR in 2026?

Google continues to hold a market-leading position and offers multiple approaches to AI-powered search.  

Google’s AI Overviews  

AI Overviews and AI Mode transform traditional queries by generating summaries instead of listing websites.  

Gemini 3.5 Flash, Google’s latest model, handles complex research tasks and in-depth questions beyond simple searches, using large-scale language models to understand user intent and provide context-driven content. 

Microsoft’s Copilot  

Evolved its Bing Chat into Copilot, creating a comprehensive AI assistant integrated across its productivity ecosystem.  

OpenAI’s ChatGPT 

Though not a traditional browser, now functions as a full search engine using natural language processing.  

Anthropic’s Claude 

Through models like Claude Fable 5 and Claude Sonnet 5, Anthropic emphasizes safety while delivering conversational answers with reduced bias. 

Brave’s AI Answer Engine, billing itself as a “non-Big Tech” search engine, provides synthesized summaries with source citations alongside traditional results, focusing on privacy as a major benefit. You.com’s YouChat lets users refine searches through interactive conversations while integrating code generation and writing assistance. Perplexity AI, once the frontrunner in AI search, pioneered source citations and delivers precise, contextually relevant answers — especially effective for factual queries.

Duck.ai by DuckDuckGo targets privacy-conscious users with GenAI summaries powered by OpenAI’s technology. Mistral, which offers open-weight models, emphasizes efficiency and accuracy for users ranging from basic information seekers to those needing in-depth analysis.

There are still others that often serve niche purposes, like Felo, which is good for multi-language searches, and Andi, which is a free AI search engine. Each platform brings unique strengths to the evolving search landscape.

Why Do Industry-Specific AI Solutions Matter?

While general AI transforms broad search, specialized tools deliver targeted value. Précis AI exemplifies this approach by building industry-specific platforms like Précis Public Relations designed exclusively for PR, communications and marketing specialists, and Willard, designed for government relations professionals. Unlike general-purpose AI, Précis understands industry terminology, workflows and deliverables. 

This specialization produces measurable results. Précis users report average time savings of 61.3% per task and 8.5 hours saved weekly. Project completion times improve by 70% to 180%, directly addressing persistent industry challenges like agency overservicing and resource optimization. 

How Should PR Pros Optimize Content for AI Search? 

Success in this new landscape requires fundamental changes to content strategy. The research on generative engine optimization is clear about what works: 

  • Lead with the answer. The first 200 words of any article should directly and completely answer the primary query, not just introduce it. 
  • Optimize for E-E-A-T. Create content that demonstrates experience, expertise, authoritativeness and trustworthiness — the factors AI models prioritize. 
  • Structure for extraction. Pages with sequential heading structures (H2 > H3 > H4) can see a 2.8x citation lift compared to unstructured equivalents. 
  • Go multi-platform. ChatGPT is the most popular, but Perplexity, Gemini and Claude each have different citation preferences — optimize for all of them, plus off-site presence on LinkedIn, Reddit, YouTube and review sites. 
  • Keep content fresh. For commercial and evaluation-stage queries, 83% of AI citations came from pages updated within the past 12 months, with more than 60% refreshed within the last six months. 
  • Nail the technical basics. Ensure fast loading, mobile optimization and HTTPS. 

The Path Forward 

The transformation accelerates daily as more users choose AI-powered answers over traditional search results. PR professionals must evolve from link-builders to trusted information sources that AI systems recognize and cite. This means creating genuinely helpful content, building relationships with industry-specific AI platforms and measuring success through new metrics beyond clicks and impressions.  

How to Measure Search Engine Optimization Success in 2026 

With the rapidly changing SEO landscape, PR pros should track the following AI-native metrics closely for the remainder of 2026: 

  • Mention Rate: the percentage of AI answers that mention your brand. 
  • Citation Rate: the percentage that include a clickable link to your domain. 
  • Position: where your brand appears when it is cited. 

There are plenty of ways to do those measurements, including manually performing searches to see what the results are. But, if you want to reduce the time it takes to track your presence across so many AI search engines, you can use tools like Semrush, Ahrefs’ Brand Radar, GUZU, Pixis or countless other emerging and established applications in the search space. 

The GEO game is changing so quickly that in six months or a year, we may see frontier AI applications become foundational, just as the models we now consider foundational were considered industry-changing just a few months or years ago. 

Download the Invisible Influencers report from Précis AI to discover comprehensive strategies for thriving in the AI-powered search era. Learn how to optimize content for AI comprehension, build authority in the new landscape and leverage specialized tools designed for PR professionals. 

Now that search engines answer questions directly, public relations pros must rethink everything about content visibility. In the first four months of 2026, roughly 68% of Google searches ended without a click (up from about 61% in 2024). Thanks to AI features, instant answers, UI elements that keep searchers in the results and shifting user preferences, Google is becoming a walled garden.  

Google’s E-E-A-T framework has become the key to staying relevant in this new landscape. E-E-A-T – Experience, Expertise, Authoritativeness and Trustworthiness – can help PR pros evaluate which content AI systems will trust and share. These signals help search engines separate quality information from noise. Furthermore, mastering them means the difference between being heard and being invisible. 

Why E-E-A-T Matters More Than Ever 

Our Invisible Influencers report reveals a startling truth: web traffic is vanishing as AI generates direct answers. Traditional SEO tactics that once guaranteed visibility? They’re losing power fast. Instead, content must prove its worth through real experience and genuine expertise. 

Notably, Google’s March 2026 Core Update became the most volatile update in its history. Sites that added structured author pages with verifiable credentials, industry affiliations and byline consistency saw measurable ranking improvements within weeks of the update. 

Experience means showing actual results, not just claiming success. Share specific campaign metrics, document real challenges and highlight lessons learned. Additionally, AI systems can spot generic content immediately, so in other words, they want proof you’ve done the work. 

Building Authority in the AI Age 

Expertise requires deep knowledge that only comes from years of doing. Use industry terms correctly, address complex topics thoroughly and provide insights newcomers couldn’t offer. Meanwhile, showcase your team’s credentials, certifications and speaking engagements. 

Authority comes from recognition by your peers. When industry leaders cite your work and share your insights, AI takes notice. Therefore, pursue quality backlinks, contribute to respected publications and participate actively in professional organizations. 

Trust: The Foundation of Everything 

Above all, trustworthiness anchors the entire E-E-A-T framework. The Invisible Influencers report emphasizes that trust outweighs all other factors. Be transparent about client relationships, verify every claim and maintain secure websites with clear contact information. 

Adapting Your PR Strategy 

The report shows that AI prioritizes well-structured content with clear headings and direct answers. Consequently, PR teams must create materials that AI can easily process and summarize. Focus on becoming the authoritative source that AI references when answering industry questions. 

This transformation demands new thinking. Rather than chasing keywords, build genuine expertise that AI recognizes and rewards. The payoff? Your content reaches highly targeted audiences at the exact moment they need answers. 

The Invisible Influencers report provides the complete roadmap for thriving as search evolves beyond traditional SEO. Download it today to discover comprehensive strategies for maintaining client visibility in the AI-driven search landscape. Because in this new era, understanding E-E-A-T isn’t optional—it’s essential for PR survival.