
Artificial intelligence (AI or later named SI) is now part of most marketing departments, newsrooms and ad agencies. Over the past year, the research has caught up with the hype, and it shows us two things. First, AI can write, design and persuade at a level that rivals skilled professionals. It can also deceive people at a scale and quality that most of us cannot detect.
Here is what the evidence says about both sides, and what I think leaders and communicators should do about it. To find quality information, I searched peer-reviewed studies from 2025 and 2026, recent industry surveys, regulatory documents and court records and here is what I learned.
AI is now standard equipment in marketing
Adoption is no longer the question. McKinsey’s global survey, published in November 2025, found that 88 percent of respondents said their organizations regularly use AI in at least one business function, up from 78 percent a year earlier. Marketing and sales, along with IT, have consistently been the functions where AI is used most, and marketing and sales use cases are among the most common sources of reported revenue gains [1]. The same survey also contains a reality check: only 39 percent of respondents reported any effect on enterprise-level profit (EBIT) from AI [1].
Salesforce’s State of Marketing report, published in February 2026 and based on nearly 4,500 marketing decision makers, found that 75 percent of marketers have adopted AI and that personalizing content is their top use. Yet 84 percent admitted they still run generic campaigns, mostly because their customer data is scattered across systems [2]. In advertising, the Interactive Advertising Bureau (IAB) reported in January 2026 that 83 percent of ad executives say their company uses AI in the creative process, up from 60 percent in 2024, and that cost efficiency is now the top benefit they cite [3].
Almost everyone is using AI, but far fewer are using it well.
The power: AI can persuade, and AI ads can perform
The most striking research of the past year concerns persuasion.
In a preregistered study in Nature Human Behaviour, GPT-4 debated 900 participants. When it had basic demographic information about its opponent, it was more persuasive than a human debater 64.4 percent of the time in debates where the two were not equally persuasive. Without that information, it performed about as well as people [4].
Two larger studies published in December 2025 went further. In Science, researchers from Oxford, the UK AI Security Institute and partner universities tested AI persuasion with nearly 77,000 UK participants. Training and prompting techniques raised persuasiveness far more than using a bigger model, and the most persuasive responses were packed with fact-checkable claims. The troubling trade-off is worth noting: the techniques that made AI more persuasive tended to make it less accurate [5]. In Nature, an MIT-led team found that conversations with an AI advocating for a candidate in the 2024 US, 2025 Canadian and 2025 Polish elections shifted preferences more than video ads typically do, though not all of the AI’s facts were accurate [6].
For marketers, the lesson is that conversation plus personal “seemingly” fact based data is a powerful combination, but accuracy does not come along automatically.
The advertising evidence is more mixed. A Marketing Science study published in October 2026 compared 4,633 pairs of AI-generated and human-made ad images on the Taboola network, run by the same advertisers in identical campaigns, across more than 369 million impressions. On average, the AI images had no click-through disadvantage, and AI images that did not look like AI outperformed human-made ones [7]. A peer-reviewed paper at the 2026 ISMS Marketing Science Conference reported a randomized field experiment with more than 150 video ads. AI-generated videos earned higher click-through rates and retention but were less likely to produce conversions [8].
AI is very good at grabbing attention but not what usually closes the sale.
The audience is moving to answer engines
AI is also changing how people find information. Pew Research Center analyzed the browsing data of 900 US adults and found that when a Google search showed an AI summary, users clicked a traditional result in 8 percent of visits, compared with 15 percent when no summary appeared. Only 1 percent of visits included a click on a link inside the summary itself [9]. That study used March 2025 data, older than my other sources, but it remains the clearest independent measurement of the effect.
More recent data point the same way. The Reuters Institute’s January 2026 trends report, using Chartbeat data, found that Google search traffic to more than 2,500 news sites fell 33 percent globally and 38 percent in the US between November 2024 and November 2025, though it is not clear how much is due to AI overviews. News executives expect search traffic to fall more than 40 percent over the next three years [10]. The Reuters Institute’s Digital News Report 2026, published in June, found that weekly use of AI chatbots for news rose from 7 to 10 percent. Across 27 markets where the question was asked, just 4 percent of respondents said they always or often click through to sources from AI, compared with 19 percent from search [11].
I wrote last month about Google’s new option to opt out of AI answers. The bigger issue is that fewer people will reach your website by default, and original, genuinely useful content has the best chance of being cited and clicked.
The deception: fraud, fakes and slop
In November 2026, federal prosecutors in New York unsealed charges against three men accused of using AI to impersonate the voices and images of company CEOs and lawyers at major law firms. Prosecutors allege the scheme took at least $129 million from businesses between July 2024 and February 2026, including a UK fintech company that lost almost $11.6 million [12]. These are allegations, and the defendants are presumed innocent, but the method is very familiar: a convincing executive, an urgent confidential deal and an employee who can wire money. Seen that, right?
The FBI’s Internet Crime Complaint Center added a dedicated AI section to its annual report this year. For 2025, it recorded 22,364 complaints involving AI, with more than $893 million in adjusted losses. Investment scams with a reported AI connection accounted for more than $632 million, and business email compromise involving AI cost businesses more than $30 million [13].
Can people spot the fakes? Recent research says, mostly, no.
- In a study published in Acoustics in June 2026, 95 listeners tried to tell real voices from voice clones played through simulated phone networks. Overall accuracy was 54.8 percent, barely better than a coin flip [14].
- At the ACM CHI 2026 conference, researchers reported that 195 participants were less accurate judging deepfake videos than real ones, yet felt just as confident about both [15].
- In a January 2026 study in Cognitive Research: Principles and Implications, people classified real and deepfake still images at chance level [16].
- In a 2025 conference study, 151 consumers trying to separate genuine restaurant reviews from ChatGPT-written fakes were right 53.2 percent of the time, and were worst at catching fake negative reviews [17].
In a Pew survey of more than 5,000 US adults, 76 percent said it is extremely or very important to be able to tell whether pictures, videos and text were made by AI or people, but 53 percent said they are not too or not at all confident they can do it [18].
Then there is the flood of low-quality synthetic content called AI slop. The Reuters Institute notes that TikTok says it hosts more than one billion AI videos, while fewer than 1 percent of news images and videos carry C2PA provenance metadata, the standard designed to show where content came from [10].
Fake reviews are an old problem that AI makes cheaper. The FTC’s rule on consumer reviews and testimonials took effect in October 2024, and the agency stated that AI-generated fake reviews are covered [26]. In 2026, the agency has used it against a home services operator accused of thousands of fake listings and five-star reviews, and a supplement company whose employees allegedly posed as customers [19]. In October 2026, it sued two more companies over alleged fabricated reviews, including one accused of paying rideshare drivers to stage photos for fake positive reviews [20].
Disclosure: the trust paradox
If deception is the problem, disclosure seems like the obvious answer. The research says it is more complicated than that.
A 2025 experiment in the Journal of Interactive Advertising found that labeling an ad as AI-generated lowered trust in both the ad and the organization. The same study also found that disclosure made people see the use of AI as more appropriate, nudging trust back up [21]. Three experiments published in the Journal of Retailing and Consumer Services in 2025 found that AI disclosures lowered trust in service ads. However, trust recovered when AI was used only for tangible elements, such as an office or equipment, while the people shown were real [22]. A 2026 study of 762 Gen Z consumers found that AI influencers raised perceptions of manipulation whether or not they were disclosed [23].
In IAB’s 2026 study, 82 percent of ad executives believed Gen Z and Millennial consumers feel positive about AI-generated ads, but only 45 percent of those consumers actually do. Even so, 73 percent of those consumers said knowing an ad was made with AI would either increase their likelihood of buying or make no difference. Fewer than half of advertisers who use generative AI always disclose it [3].
My reading of all this: disclosure costs a little trust in the short run, but being caught hiding AI costs much more. Use AI freely behind the scenes, never let it pose as a real person, and be open whenever AI content could be mistaken for the real thing.
The rules are arriving
In the European Union, the AI Act’s transparency obligations under Article 50 have applied since August 2, 2026. Deepfakes and AI-generated text published to inform the public on matters of public interest must be clearly labeled, and people must be told when they are interacting with a chatbot [24]. The European Commission published a voluntary Code of Practice on marking and labeling AI-generated content in June 2026, including an optional EU icon. AI systems already on the market before August 2 have until December 2, 2026 to comply [25]. US businesses with European customers should pay attention.
In the United States, there is no equivalent federal labeling law, but the FTC’s review rule and its general authority over deceptive practices already cover fake testimonials and fabricated customers, including AI-generated reviews [19][20][26]. Platforms are adding their own requirements as well. YouTube, for example, now requires creators to disclose altered or synthetic content that could be mistaken for a real person, place or event [10].
What I recommend for businesses and communicators
- Use AI for speed, not for fake people. Drafting, research, transcription and personalization are where AI earns its keep. Never use it to invent customers, reviews, testimonials or endorsements. That is where the legal risk and trust damage live.
- Keep the human parts human. The service-ad research suggests a simple rule: AI can build the set, but real people should be the faces of your brand [22].
- Write a disclosure policy and follow it every time. Decide in advance when you will label AI-generated images, video, voices and chat, and be consistent. In the EU it is now the law for certain content [3][24].
- Fact-check everything AI produces. Persuasive AI is not always accurate AI [5][6]. A human should verify every claim before it goes out under your name.
- Measure conversions, not just clicks. AI creative can win attention and still lose the sale [8]. Judge it by inquiries, sales and retention.
- Prepare for impersonation. Require call-back verification on any request to move money, no matter who appears to be asking, and monitor for fake accounts and ads using your brand or executives’ likeness [12][13].
- Be worth the click. Publishers surveyed by the Reuters Institute plan to invest in original reporting, analysis and human stories rather than generic content that AI can summarize [10]. Businesses should do the same with their own expertise.
- Lead from the top. In McKinsey’s survey, the organizations getting the most value from AI were more likely to have defined processes for when AI output needs human review [1]. Leaders need to model responsible use, a theme I explored in CEOs Think Employees Trust Them on AI. Most Employees Don’t.
The bottom line
AI has given marketers and media organizations remarkable power to create and to persuade. Used by the wrong people, the same power deceives customers and erodes trust. The past year’s research is clear that most people cannot reliably tell real from fake. That puts the responsibility on those of us who create content to be honest about how we make it.
If you would like help building an AI use and disclosure policy for your organization, or a communication plan that uses these tools without sacrificing trust, schedule a consultation with me.
References
Peer-reviewed sources are marked [Peer-reviewed]. All other sources are industry research, government documents, legal analysis or news reports.
- Singla, A., Sukharevsky, A., Hall, B., Yee, L., Chui, M., & Balakrishnan, T. (2025, November 5). The state of AI in 2025: Agents, innovation, and transformation. McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2025
- Salesforce. (2026, February 19). 75% of marketers have adopted AI, yet still use it to send generic campaigns [State of Marketing, 10th ed.]. https://www.salesforce.com/news/stories/state-of-marketing-2026/
- Williamson, D. A., Koch, J., & Giegerich, C. (2026, January 15). The AI ad gap widens. Interactive Advertising Bureau. https://www.iab.com/insights/the-ai-gap-widens/
- [Peer-reviewed] Salvi, F., Horta Ribeiro, M., Gallotti, R., & West, R. (2025, May 19). On the conversational persuasiveness of GPT-4. Nature Human Behaviour, 9(8), 1645–1653. https://doi.org/10.1038/s41562-025-02194-6
- [Peer-reviewed] Hackenburg, K., Tappin, B. M., Hewitt, L., Saunders, E., Black, S., Lin, H., Fist, C., Margetts, H., et al. (2025, December 4). The levers of political persuasion with conversational artificial intelligence. Science, 390(6777), eaea3884. https://doi.org/10.1126/science.aea3884
- [Peer-reviewed] Lin, H., Czarnek, G., Lewis, B., White, J. P., Berinsky, A. J., Costello, T., Pennycook, G., & Rand, D. G. (2025, December 4). Persuading voters using human–artificial intelligence dialogues. Nature, 648(8093), 394–401. https://doi.org/10.1038/s41586-025-09771-9
- [Peer-reviewed] Exner, Y., Hartmann, J., Ding, Z., Zhang, S., & Netzer, O. (2026, October 19). AI in disguise: Quasi-experimental analysis of a large-scale deployment of AI-generated ads. Marketing Science. Advance online publication. https://doi.org/10.1287/mksc.2025.0547
- [Peer-reviewed conference paper] Li, H., & Li, Q. (2026, June 12). Attention without conversion? AI-generated versus human-created video ad creatives in a randomized field experiment [Conference paper]. 48th Annual ISMS Marketing Science Conference, Carcavelos, Portugal. https://scholars.hkbu.edu.hk/en/publications/attention-without-conversion-ai-generated-versus-human-created-vi/
- Chapekis, A., & Lieb, A. (2025, July 22). Google users are less likely to click on links when an AI summary appears in the results. Pew Research Center. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- Newman, N. (2026, January 12). Journalism, media, and technology trends and predictions 2026. Reuters Institute for the Study of Journalism. https://reutersinstitute.politics.ox.ac.uk/journalism-media-and-technology-trends-and-predictions-2026
- Reuters Institute for the Study of Journalism. (2026, June 16). Digital news report 2026. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/dnr-executive-summary
- Kapko, M. (2026, November 6). Authorities accuse trio of using AI deepfakes to defraud $129M from businesses. CyberScoop. https://cyberscoop.com/ai-deepfakes-impersonate-ceo-lawyer-defraud-companies/
- Federal Bureau of Investigation, Internet Crime Complaint Center. (2026, April). 2025 IC3 annual report. https://www.ic3.gov/AnnualReport/Reports/2025_IC3Report.pdf
- [Peer-reviewed] Warzych, J., Łuczyński, M., & Klink, J. (2026, June 17). Human detection of voice-cloned speech under GSM, VoLTE and VoIP conditions. Acoustics, 8(2), 41. https://doi.org/10.3390/acoustics8020041
- [Peer-reviewed] Chen, C., & Goh, D. (2026, April 13). Seeing, hearing, and knowing together: Multimodal strategies in deepfake videos detection. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (pp. 1–12). ACM. https://doi.org/10.1145/3772318.3790480
- [Peer-reviewed] Pehlivanoglu, D., Zhu, M., Zhen, J., et al. (2026, January 7). Is this real? Susceptibility to deepfakes in machines and humans. Cognitive Research: Principles and Implications, 11, Article 3. https://doi.org/10.1186/s41235-025-00700-y
- [Peer-reviewed conference paper] Fröhnel, K., Santelmann, B., & Zarnekow, R. (2025, January). Genuine or fake? Explaining consumers’ perception and detection of AI-generated fake reviews. In Proceedings of the 58th Hawaii International Conference on System Sciences (pp. 4219–4228). https://hdl.handle.net/10125/109350
- Kennedy, B., Yam, E., Kikuchi, E., Pula, I., & Fuentes, J. (2025, September 17). How Americans view AI and its impact on people and society. Pew Research Center. https://www.pewresearch.org/science/2025/09/17/how-americans-view-ai-and-its-impact-on-people-and-society/
- Atleson, M., Boyle, B., & Stewart, J. (2026, July 27). FTC’s 2026 enforcement approach to fake reviews takes shape: Takeaways for companies. DLA Piper. https://www.dlapiper.com/en/insights/publications/2026/07/ftcs-2026-enforcement-approach-to-fake-reviews-takes-shape-takeaways-for-companies
- Atleson, M., Boyle, B., & Stewart, J. (2026, November 2). FTC’s actions against fake reviews in GovAssist and Digital Forensics: Top points. DLA Piper. https://www.dlapiper.com/en/insights/publications/2026/11/ftcs-actions-against-fake-reviews
- [Peer-reviewed] Koning, B., & Voorveld, H. A. M. (2025, July 3). Disclaimer! This content is AI-generated: How AI-disclosures influence trust in advertisements and organizations. Journal of Interactive Advertising, 25(3), 240–253. https://doi.org/10.1080/15252019.2025.2554149
- [Peer-reviewed] Grigsby, J. L., Michelsen, M., & Zamudio, C. (2025, May). Service ads in the era of generative AI: Disclosures, trust, and intangibility. Journal of Retailing and Consumer Services, 84, 104231. https://doi.org/10.1016/j.jretconser.2025.104231
- [Peer-reviewed] Yıldırım, E., & Dursun, F. (2026, June 21). Disclosure matters: Perceived manipulation, perceived ethics, and purchase intention toward AI influencers in social media marketing. Journal of Theoretical and Applied Electronic Commerce Research, 21(6), 194. https://doi.org/10.3390/jtaer21060194
- European Commission. (2026, June 10). Commission publishes Code of Practice on marking and labelling AI-generated content [Press release]. https://digital-strategy.ec.europa.eu/en/news/commission-publishes-code-practice-marking-and-labelling-ai-generated-content
- European Commission. (2026, June). Code of Practice on transparency of AI-generated content: Questions and answers. https://digital-strategy.ec.europa.eu/en/faqs/code-practice-transparency-ai-generated-content
- Federal Trade Commission. (2024, August 22). Trade regulation rule on the use of consumer reviews and testimonials. Federal Register, 89(163). https://www.govinfo.gov/content/pkg/FR-2024-08-22/html/2024-18519.htm (Older than the 2025 to 2026 window; included as foundational context because it is the rule currently in force.)