Last month, the Chicago’s American Marketing Association (AMA) hosted a diverse panel of experts to explore top trends in marketing for 2025. Versta Research was honored to contribute, sharing our perspective on the evolving landscape of marketing research. While we cannot presume to forecast the future for our entire industry, we focused on six (6) developments we’re actively encountering, grappling with, and innovating around—trends that are already significantly shaping our work in the year ahead.

1. Responding to AI Threats

Marketing research has been facing a mini-crisis of data quality for the past decade. Even before the rise of AI, there were studies—both published and unpublished—based on obviously erroneous data. Many companies, both clients and vendors, lacked the time, budget, and expertise needed to identify and eliminate fraudulent data.
Well guess what? It is now a full-blown industry crisis as artificial intelligence generates fraudulent data that increasingly mimics authentic responses. The biggest problem stems from widespread use of online panels to source research respondents. These respondents are paid for their time and effort, as they should be. But it has invited massive amounts of fraud from survey respondent sweat shops.

We are now facing a full-blown industry crisis as artificial intelligence generates fraudulent data

Versta Research has always reviewed and cleaned out bad respondents, but AI is making our jobs much harder, and for those of us who do have the resources to ferret out fraud, it is putting our skills and ingenuity to the test. Suddenly, tell-tale inconsistencies we used to flag fraudulent responses are disappearing. Responses to open ends are no longer full of suspicious syntax. At first the responses sounded like Shakespeare, so perfect was the grammar. But AI tools have evolved even in the last few months and now the fraudulent responses to open ends sound frighteningly like the real consumers in our surveys.

Clients are starting to insist on measures to ensure quality data that are coming from authentic survey respondents. (Thank goodness they have finally noticed!) We have one client who now requires documentation of data cleaning processes and outcomes so that problems with fraud can be measured and tracked over time.

Also of note, Proctor and Gamble recently announced that, as of June 2025, any research vendor who supplies respondents for their research must be ISO 20252 certified, an international standard for market, opinion, and social research, that ensures consistent quality in research processes. Having been ISO certified ourselves (to ISO 27001, which focuses on information and data security) we can attest that this is a huge hurdle for research companies, reflecting the deeply serious nature of the problem being faced by some of the biggest sponsors of consumer research.

2. Navigating Use of AI Tools

While the biggest issue in AI for market research currently is the threat it poses to our data, there is also huge potential for AI benefiting our work. Alas, most of the current tools are rudimentary at best, and many of them are downright terrible. Overall, we are not yet seeing a consistent payoff in what these tools can do.

But there is promise. Chat GPT is becoming extremely useful for brainstorming, generating lists, categories, perspectives, and vocabulary. The ideas generated through Chat GPT make their way into our proposals and questionnaires, and we absolutely see improvement and value. Microsoft CoPilot, on the other hand, has been a dismal and annoying failure for us. Also note: Chat GPT absolutely cannot write questionnaires, discussion guides, useful summaries, or reports, unless you want vacuous templates that look and “sound” like the real thing.

Specific to market research, there are emerging tools that promise to code open-ends, conduct interviews and focus groups, analyze data, and write code. However, most providers to date are refusing to demonstrate replicability with our own data and previous work. Others lack the secure practices needed to meet ISO 27001 confidentiality and data protection standards. These shortcomings are significant barriers, even as we actively trial some of their solutions.

There is also quite a bit of experimentation around generating synthetic respondents from which to conduct surveys and focus groups. Our analysis so far is that these efforts generate research that looks amazingly real, but is, in fact, entirely fake and misleading.

3. Dissipation of the Research Function

A third trend we are seeing in our work is that corporate and organizational research is increasingly being integrated into other professional areas and business units. Organizations are dismantling centralized research teams and assigning researchers to other groups. Likewise, research vendors are pivoting to position themselves as “strategy” firms that resemble marketing agencies more than traditional research specialists. This doesn’t mean research is less important. Rather, it reflects efforts to embed it more deeply into the work it supports and broaden the perspective and skill set of those who are doing research.

Organizations are dismantling centralized research teams and assigning researchers to other groups

This shift has both upsides and downsides. On the downside, organizations are losing their deep, specialized, expertise. Research requires unique skills that take years to master, such as statistical methods, sampling strategies, probabilistic reasoning, and more. These are not the skills that marketing people or business strategists have, or even should have. And with decentralization, organizations are losing institutional knowledge and their ability to share insights across the enterprise.

On the upside, embedding research directly into business units brings researchers closer to decision-makers. This proximity makes research more immediate and useful while allowing researchers to balance technical expertise with strategic thinking. It’s an adjustment, but one that offers valuable opportunities for growth, new perspectives, and stronger connections to the people who use and rely on our insights.

4. The End of Metrics Mania

A huge downside to advances in technology over the last 20 years has been our increasing ability to measure everything everywhere all at once (and cheaply, too). Research sponsors believed it would help them, and so measure we did. Everything. Everywhere. All at once. Has it been valuable? In some ways, yes. It has allowed us to develop methods and new technologies that can now be applied more strategically for things that matter. We’ve stretched our abilities and understanding and can be more flexible in how we approach research.

But for organizations focused on ticking off KPIs, NPS scores, and compliance boxes, the results haven’t delivered much. The more we measure with precision, the closer we get to replicating the messy, ambiguous, and contradictory nature of reality itself. What good does that do? (See A Tale of Data That Got Too Big.) Thankfully, we are now seeing more sponsors stepping back, redefining what truly matters, and using research strategically to address important questions—not just to churn out metrics. We are seeing a refreshing and renewed hunger for insight over measurement.

One additional sign of this shift is the waning success of companies that migrated away from doing research to offering technology platforms. We see many of these firms struggling, and our research colleagues who joined them are now moving into other industries. All of this could signal the end of metrics mania—and a fresh focus on the power of meaningful research.

5. Back to Research Basics

A corollary to ending the focus on metrics is this: back to research basics. We are seeing clients, research sponsors, and business units asking more fundamental questions, and we are seeing researchers themselves reasserting the fundamentals of truly good (and insightful) research. This means focusing on meaningful and answerable hypotheses, laying out methods and protocols specifically to test those hypotheses, conducting rigorous sampling to ensure data validity and reliability, thinking strategically about analysis, and building reports that answer questions and lay out next steps.

Two recent examples highlight this back-to-basics trend. The first comes from political polling. After a disastrous 2016 election prediction that Hillary Clinton’s chances of winning were 89%, the industry took a hard look inward, and thankfully we are seeing fewer statistical aggregation models being used than in the past. By 2020, polling had improved, and by 2024, results were even stronger.

Out with flashy models and “credibility intervals;” in with old fashioned sampling, weighting, and survey design. Adhering to the foundations of survey research that have held us in good stead for many decades is paying off.

Another example comes from a recent meeting with a research sponsor. While reviewing a complex statistical analysis we performed, the sponsor posed a specific question about a subset of their customers. Instead of proposing deeper analysis, the head of research simply said, “Let’s call 10 of them and conduct interviews. That will give us the answer.” It was a refreshing moment of clarity: have a question, ask the right people in the right way, and you’ll get the answers you need.

6. Full Normalization of Accessible Surveys

The last trend of 2025 to highlight harks back to our work in 2016 when we were among the first market research firms to start fielding fully accessible surveys. This was achieved by partnering with an online platform to custom build and code a version of their software to meet WCAG 2.1 Level A and AA success criteria (“accessibility criteria”) so that people who use screen readers or other assistive devices have full and easy access to all survey content.

Research sponsors have been slow to demand this type of accessibility, and only recently have we started to see growing interest. One of our recent MSAs specifies that any survey of customers must be fully accessible and audited (tested for accessibility) by an external firm. Achieving this is no small feat. Most survey platforms claim to meet accessibility standards, but in our experience they fall short (and will fail an audit) without additional intervention, custom coding, and expertise.

One other indicator of this trend: This past year we participated for the first time in a survey that was conducted fully in American Sign Language (ASL). Many deaf and hard of hearing people have difficulty with online surveys because they are typically offered in written English, which is often a second (and more difficult) language for native users of ASL. Online video tools now make it entirely possible for surveys to be administered in ASL.

Here is a video clip from the survey we recently took, where the interviewer is communicating the answer scale from very likely to very unlikely. Even if you don’t know ASL, you will probably love the visual communication of this survey scale, and it will open your eyes to a new world of survey and research possibilities for 2025.