PR shapes public perception of AI by replacing abstract claims with named, verifiable uses, explaining how a system works and where it fails, and putting accountable people in front of journalists. The task is hard. A Pew Research Center survey conducted June 22 to 28, 2026 found that 52% of U.S. adults were more concerned than excited about AI, while 9% were more excited than concerned.
Last updated November 19, 2026.
Key takeaways
- Pew found 52% of U.S. adults more concerned than excited about AI in June 2026, up from 37% in 2021.
- Adults under 30 now lead the shift: 55% are more concerned than excited, the first majority for that age group.
- 71% of U.S. adults expect AI to lead to fewer jobs over the next 20 years, so employment is the first worry a message must address.
- A peer-reviewed 2025 experiment found that transparency about an AI system weakened the link between general distrust of AI and distrust of the company behind it.
- The working method is to name the deployment, show the human review step, publish the limits, and prepare a spokesperson before an incident.
What does the public currently think about AI?
The public is wary of AI in daily life even as it uses AI tools. Pew's survey of 3,488 U.S. adults, conducted June 22 to 28, 2026, found 52% more concerned than excited, 37% equally concerned and excited, and 9% more excited than concerned.
| Survey year | More concerned than excited | Equally concerned and excited | More excited than concerned |
|---|---|---|---|
| 2021 | 37% | 45% | 18% |
| 2023 | 52% | 36% | 10% |
| 2025 | 50% | 38% | 10% |
| 2026 | 52% | 37% | 9% |
Concern has risen from 37% in 2021, and excitement has fallen by half, from 18% to 9%. The climb continues among adults under 30: 55% now say they are more concerned than excited, up from 31% in 2021, and Pew reports it is the first time a majority of that age group has said so.
Jobs sit behind much of that worry. Pew found that 71% of adults think AI will lead to fewer jobs in the U.S. over the next 20 years, up from 64% in 2024, while 5% expect more jobs.
Those numbers change the job. An AI company cannot assume goodwill, so every claim needs evidence attached and every message needs to answer a worry the audience already has. The guide to ethical AI in PR covers the transparency standards behind that approach.
How should PR counter science-fiction stereotypes about AI?
PR counters stereotypes by showing named, real deployments in place of generic promises. Fear of AI draws on fiction, from HAL 9000 in 2001: A Space Odyssey to the machines of Terminator, and a concrete example of AI doing a bounded, useful job gives audiences something real to weigh against those images.
| Audience worry | What PR shows instead | Evidence to attach |
|---|---|---|
| AI will act on its own and escape control | A named deployment where people review and approve every output | A description of the human review step and who performs it |
| AI will replace workers | A named team that uses AI for one task and keeps its staff | The task, the tool, and the people who still own the decision |
| AI is a black box | A plain-language account of what data goes in and what comes out | A published description of inputs, outputs, and known limits |
The jobs row needs the most care. With 71% of adults expecting fewer jobs, a message that says AI assists workers without naming the task, the tool, and the person who keeps the final decision will read as a dodge.
Pick applications where the benefit is measurable and the audience can check it, such as an AI-assisted screening tool tested in a clinical study or a grid-optimization system run by a utility. Name the deployment, the measure, and the source.
How should an AI company explain what its product does?
An AI company should explain its product by defining terms once, in everyday words, and then describing what the system takes in, what it produces, and where a person steps in. Most audiences know AI only through broad labels such as "machine learning" and "deep learning," so a message built on those labels explains nothing.
Several formats carry that explanation. A press release that defines each technical term on first use, an interview with a named researcher who answers a reporter's skeptical question directly, an explainer video, and an infographic that follows one piece of data through the system all work. Each should answer the same three questions: what the system does, what it cannot do, and who is accountable when it is wrong.
A simple test applies before anything is published. A reporter who has never seen the product should be able to repeat the explanation in one sentence, and a customer should be able to find the human review step without asking.
What does transparency look like in practice for AI developers?
Transparency in practice means publishing documents the public can read and check: how the company collects and uses data, how the model was tested for bias, what its known limits are, and how a person can contest an automated result. A promise to be responsible without those documents gives skeptics nothing to verify.
The evidence that this works is specific. A peer-reviewed experiment by Keonyoung Park and Ho Young Yoon, published in Humanities and Social Sciences Communications in 2025, found that high AI algorithm transparency weakened the link between a general negative attitude toward AI and distrust of the company behind it, especially when readers cared a great deal about the issue.
How much explanation people want depends on how the product performs. A 2026 experiment reported by Penn State in Computers in Human Behavior, run on a fictional dating site, found that a simple explanation was enough when the system overdelivered, while users wanted a detailed one when it underdelivered. A company should therefore keep a short standard explanation ready for normal operation and a detailed one ready for the day the product falls short.
| Document to publish | What a reader learns | Who should be able to check it |
|---|---|---|
| Data use statement | What data the system collects, keeps, and shares | Customers, regulators, privacy researchers |
| Bias testing summary | Which groups the model was tested on and what the results showed | Journalists, advocacy groups, academics |
| Known limits page | Tasks the system handles poorly or should not be used for | Customers and reporters |
| Appeal process | How a person contests an automated result and who reviews it | Affected users and regulators |
Ethical guidelines for developers belong in the same place as the product information, written for a general reader and updated when the product changes. Discriminatory outputs and privacy breaches are the two concerns audiences raise most, so the guidelines should address both by name and describe the process used to catch them.
How should AI companies work with journalists?
AI companies should work with journalists by pitching specific, checkable stories and by preparing their researchers and engineers to explain those stories accurately. Reporters covering AI are asked to separate real capability from marketing, so the companies that earn balanced coverage supply evidence on request.
| What a reporter needs | What the company supplies |
|---|---|
| A claim they can verify | Documentation, test results, and a named person who stands behind them |
| A source who can explain it plainly | A researcher or engineer who has completed media training |
| The other side of the story | A direct statement on known risks, limits, and open questions |
Build relationships before a launch, not during one. Regular story pitches and expert commentary on AI news position a company as a reliable source, which is the foundation of the work described in the guide to media relations.
Which stakeholders need their own AI messages?
Four groups need separate messages: the general public, policymakers and regulators, industry customers, and advocacy organizations. Each starts from a different concern, and a single statement written for all four usually answers none of them.
| Audience | Main concern | Channel that reaches it |
|---|---|---|
| General public | Privacy, jobs, and whether the product can be trusted | News coverage, explainers, and social video |
| Policymakers and regulators | Safety, bias, and accountability | Briefings, public comment filings, and expert testimony |
| Industry customers | Reliability, data handling, and return on investment | Trade media, case studies, and direct account contact |
| Advocacy groups | Harm to specific communities | Direct meetings and published responses to their findings |
The same facts should appear in every message. Only the emphasis and the evidence change.
Concern also differs by political group. Pew's September 2026 analysis of the same June survey found that 56% of Democrats were more concerned than excited about AI in daily life, compared with about half of Republicans. A public message written for one partisan audience will miss a large share of the other.
What belongs in an AI company's crisis plan?
An AI company's crisis plan should name the incidents most likely to hit it, such as a biased or harmful output, a data privacy breach, or a model failure that affects customers, and assign a spokesperson and an approval path for each. A swift, specific response limits the damage when concern becomes news.
The plan should also state when the company will pause a feature, how it will notify affected users, and who will explain what happened in technical terms. The first holding statement should be readable on its own, because reporters quote it without the surrounding thread. The crisis PR guide sets out the response sequence, including the section on B2B tech companies, and the study Nvidia crisis in AI answers shows what four AI engines said on the first day of a real incident.
Frequently asked questions
Why does public perception matter for an AI company?
Customers, regulators, and employees all decide whether to trust a product partly on what they have read about it, and skepticism is widespread, with 52% of U.S. adults in Pew's June 2026 survey more concerned than excited.
What is the best way to counter fears about AI?
Show a named, real deployment with a measurable result and a visible human review step, and publish the evidence.
Should an AI company discuss its product's limits publicly?
Yes. A stated limit gives journalists and customers a reason to believe the rest of the claims, and the 2025 transparency experiment in Humanities and Social Sciences Communications found that transparency about how an AI system works weakened the link between general distrust of AI and distrust of the company behind it.
How do younger adults view AI?
Younger adults are now among the most wary. Pew's 2026 survey found that 55% of adults under 30 are more concerned than excited about AI, and 73% of them expect AI to lead to fewer jobs.
How is this different from using AI inside a PR team?
This guide covers communicating about AI products. For how communications teams use AI in their own work, see how AI is transforming corporate communications.
How often should an AI company update its public explanations?
Every time the product, the data it uses, or its known limits change, and at least once a year. Published statements that no longer match the product are the first thing a skeptical reporter checks.
What should an AI company do first?
An AI company should write down what its product does and cannot do in plain words, publish the evidence behind its claims, and prepare a named spokesperson for the hardest question a reporter is likely to ask. Five steps put that into practice:
- Write a one-page plain-language description of the product: inputs, outputs, and the point where a person reviews the result.
- Publish a known limits page and a data use statement, and date both.
- Choose one named deployment with a measurable result and prepare it as the lead example for reporters.
- Train a researcher or engineer to answer a skeptical question on the record, starting with jobs, privacy, and bias.
- Draft crisis statements for the three most likely incidents, with a spokesperson and an approval path for each.
Companies that want outside support can start with the AI, digital marketing, and PR practice or the technology PR practice.






