What Did This Quarter's Research Tools Actually Replace?

Edu here. The quarter closes today, and this week I did something slightly tedious. Every Thursday a research-tech company called Emporia publishes a digest of the market research industry called briefd, and I went through all eleven issues that cover July to September and wrote down every product launch in them.

My first instinct was to write the usual roundup, the one where AI is eating research and here are the ten tools to watch. Then I actually sorted the list by what each launch replaces, meaning the thing a DTC brand does today (or pays someone to do) that the new tool takes over. That turned out to be a more useful question, and the answers are more lopsided than I expected.

What shipped, counted

Those eleven issues ran 260 items. Most of them were people changing jobs and opinion pieces, which I dropped. What was left, after merging the launches that showed up in two weeks running, was 69 products or features.

A few limits on that count, because I'd want to know them if someone else showed me this chart. briefd leans towards enterprise research and the UK trade press, so a Shopify app announced with a blog post and a tweet would never appear in it. The count covers eleven of the quarter's thirteen weeks, because one late-July issue returns a 404 and the issue for the last days of September isn't out yet (I also left out the first July issue, since it mostly covered the end of June). The digest also says at the top that it's AI-generated and "may occasionally make mistakes", so I used it as an index and checked every launch I name below against the company's own announcement or the original trade story. Some of the categories are judgment calls (is a fraud check a research tool? I said yes). Move a handful of launches between buckets and the shape stays the same.

The biggest group replaces an export

25 of the 69 launches, more than a third, do one thing: they let you ask questions in plain language about data you already have. Mostly that means connecting through the Model Context Protocol, the standard that lets ChatGPT or Claude read a company's database. Nielsen, Mintel, Euromonitor, NIQ, Circana, Rival, Voxpopme and Caplena all shipped a version of this.

For a brand under $5 million in revenue, most of these replace nothing, because they sit on top of a data license you don't hold (nobody at a twelve-person supplements brand is paying for Mintel). The exception is SurveyMonkey. On August 12, 2026 it launched a ChatGPT plugin that lets you build surveys and analyze responses inside the chat, after a Claude connector in May. If your post-purchase survey lives in SurveyMonkey, this replaces the CSV export and the pivot table. That's a real saving, maybe an hour a month.

It doesn't replace deciding which rows count, though, and that's where I spend most of my time. When I built the AI shopping trust study in August, about 10% of the comments I first pulled were actually about shopping. The rest were AI fiction, code reviews, court filings and stock picks that happened to share the vocabulary. A chat layer would have summarized all of it with total confidence.

Simulated customers replace the first cut, and the vendors say so

Twelve launches simulate customers in some form (digital twins or AI personas, mostly). The quarter's biggest funding round went this way too. Simile raised $200 million at a $2 billion valuation on July 30, five months after its Series A, to sell simulated users for marketing and product research.

What surprised me is how carefully the launches themselves limit the claim. Zappi's Synthetic Idea Pre-Screener, launched on August 27, takes 30 to 300 early ideas and ranks them in about 15 minutes. Zappi's own figure is that 88% of the ideas it ranks in the top 30% also do at least medium-well with real people. Stravito's persona builder, due for general availability in Q4, links every trait in a persona back to the research it came from. Nobody at the launch stage is claiming to replace a survey.

The independent tests this quarter explain why. Strat7 ran real surveys of 3,000 people against two synthetic data companies. The synthetic answers landed two to three points from the real headline numbers, which sounds great until you get to the rest of the paper. They quoted prices about 16% higher than real consumers, and they tracked the real change between 2025 and 2026 correctly 47% of the time, which is coin-flip territory. In August a Prolific pre-print found that giving each simulated respondent a demographic persona roughly tripled the error compared with just asking the model for an average. I wrote about the wider problem with synthetic respondents earlier this month, so I won't repeat it here.

So for a DTC brand, the honest replacement is the first cut of a long concept list, the afternoon you'd spend choosing which six of forty product names to test properly. Zappi's own page puts it better than I can: the tool is

"designed to answer which ideas merit a closer look, not to declare which ideas are proven winners."

AI interviewers replace the call you never booked

Only four launches were AI interviewers, but they're the ones I'd take most seriously if I ran a brand. On September 15, YouGov extended Voices, its AI-moderated follow-up interview, to custom research in the US, UK and Australia. Two days later Marvin launched Live Intercept, which puts an AI interviewer inside your website. Once a visitor agrees, it follows your discussion guide, asks follow-ups and lets them talk out loud or share their screen.

Customer interviews are the thing most small brands know they should do and don't, because recruiting ten customers and finding the hours is miserable. Live Intercept goes after the recruiting problem directly. Marvin's CEO said as much: customers had run thousands of AI interviews on the platform, but "recruiting the right customers was still slowing research down." For a brand with decent site traffic, this replaces the post-purchase survey question that nobody reads and the interview program nobody started. It's the one launch this quarter that takes over something a small brand does, or should be doing.

The money agrees. Listen Labs, which runs AI interviews for Microsoft and Sweetgreen among others, on about $30 million in annualized revenue, walked away from a signed $125 million round, TechCrunch reported on September 9, most likely because of acquisition talks with Salesforce at around $2 billion. Whether AI moderators match good human ones is still thin on evidence (I covered the one small comparison I could find in the post on the new ISO standard). Speed and access are the clear wins. Depth, I'm not sure yet.

Tracking AI answers replaces nothing, because the question is new

Six launches track how ChatGPT and Google's AI Mode talk about brands. Profound raised $180 million at a $1.8 billion valuation on September 15, 2026. It says its platform is built on more than two billion real user prompts.

I'd put these in their own column, because they don't take over anything a brand did before. Two years ago nobody had to ask what an answer engine says about their magnesium. Reach3's AI Listening, launched in late July, makes the useful point that the answers change with who's asking and can differ between two identical prompts, so a one-off check tells you very little. I went into why customer language matters for these answers in the post on AI visibility. The short version is that AI engines cite Reddit more than any other source, so the threads your customers write in shape what gets said about your category.

The one launch that reads what people say unprompted

One launch out of 69. Bastion Insights, an Australian agency, launched Vantage at the end of July, pairing AI analysis of online conversation with senior researchers, and it describes the product as a complement to its research rather than a replacement. I also went looking for anything Reddit itself shipped for brands this quarter, and couldn't find a thing (which says more about my search than about Reddit, maybe).

I should say the obvious thing here. Reading unprompted conversation is my job, so of course I noticed it was the empty bucket (you'd be right to discount this section a little for that). But the count is the count. The quarter's money and launches went into faster ways to ask people questions, simulated people to ask instead, and chat windows over answers already collected. Almost nothing went into listening to what customers say to each other when no brand is in the room.

I don't have a good explanation for why. Maybe Reddit's data licensing prices make it a hard product to build at scale. Maybe it's that the expensive part, deciding which 10% of a pull is relevant and what the pattern means, doesn't turn into software cleanly. I guess it could be both. I'm not sure, and I'd like to be wrong about the second one, because it would make my own work faster.

Frequently asked questions

What research tools launched in Q3 2026?

Across eleven weekly issues of the briefd industry digest covering July 3 to September 24, 2026, 69 research products or features launched. The largest group, 25, added an AI chat or Model Context Protocol layer over data the buyer already owns or licenses. Twelve simulated customers, six tracked brand visibility in AI answers, six checked survey fraud and four were AI interviewers.

Are synthetic respondents accurate enough to replace a survey?

Not for anything past a first screen. A 2026 Strat7 study against 3,000 real respondents found synthetic results sat two to three points from headline answers, but tracked real change between 2025 and 2026 correctly only 47% of the time and quoted prices about 16% higher. A Prolific pre-print found demographic personas tripled distributional error.

What is an AI-moderated interview?

An interview where software asks the questions and adapts its follow-ups to each answer, with no human moderator. YouGov launched Voices for custom research on September 15, 2026, and Marvin launched Live Intercept on September 17, 2026, which starts an interview inside a website once a visitor consents.

What does an AI visibility tool like Profound do?

It tracks how answer engines such as ChatGPT and Gemini describe and recommend a brand. Profound raised a $180 million Series D at a $1.8 billion valuation on September 15, 2026, and says its platform is built on more than two billion real user prompts. It measures a new question rather than replacing an existing research method.

Which Q3 2026 research tools can a small DTC brand actually use?

Only a handful of the 69 launches were self-serve at small-brand budgets. SurveyMonkey's ChatGPT plugin (August 12, 2026) and its self-serve GetFeedback widget are the most accessible. Marvin's Live Intercept fits a brand with website traffic. Most of the rest sit on top of enterprise data licenses a brand under $5 million in revenue does not hold.

What to do with this

If you run a small brand, look at two of these and ignore the rest for now. If your feedback already sits in SurveyMonkey, try the ChatGPT plugin and see if it saves you the monthly export. If you have the traffic, test an on-site AI interviewer against your post-purchase survey for a month and compare what comes back.

Also, treat any synthetic panel as a way to shorten a list, never as a way to pick the winner. The vendors are telling you that themselves.

All in all, Q3 was a big quarter for asking faster and a quiet one for listening. I'll run the same count at the end of Q4 and see if that changes. If you spotted a launch this list missed (especially a Shopify-side one, which briefd would never catch), send it over and I'll add it to the next count.

Sources

1. Emporia Research. briefd: Market Research & Insights, weekly issues dated July 9, July 16, July 30, August 6, August 13, August 20, August 27, September 3, September 10, September 17 and September 24, 2026. Source for the launch count; the digest states it is AI-generated. Latest issue: Link Retrieved September 30, 2026.
2. SurveyMonkey. (2026). SurveyMonkey launches plugin in ChatGPT to bring real human feedback into AI workflows, August 12, 2026. Link Retrieved September 30, 2026.
3. TechCrunch. (2026). Synthetic-user startup Simile raises $200M at $2B valuation 5 months after $100M Series A, July 30, 2026. Link Retrieved September 30, 2026.
4. Zappi. Synthetic Idea Pre-Screener, product page. Source for the 30 to 300 ideas, the 15-minute turnaround, the 88% validation figure and the quoted positioning. Link Retrieved September 30, 2026.
5. Research Live. (2026). Synthetic data survey responses "significantly different" to humans, finds study. Strat7 paper "Synthetic data: Is this as good as it gets?", source for the 3,000-person comparison, the two to three point gap, the 16% price gap and the 47% tracking rate. Link Retrieved September 30, 2026.
6. Research Live. (2026). Synthetic respondents can't replace human measurement, finds Prolific study, August 24, 2026. Covers Gordon et al., "Do Richer Personas Improve LLM Survey Simulation? A Fidelity Paradox". Link Retrieved September 30, 2026.
7. YouGov. (2026). Launching YouGov Voices: qual at scale, September 15, 2026. Link Retrieved September 30, 2026.
8. MRWeb Daily Research News. (2026). Marvin launches Live Intercept, September 17, 2026. Link Retrieved September 30, 2026.
9. TechCrunch. (2026). AI research startup Listen Labs scrubbed a $1.5B funding round for Salesforce talks, September 9, 2026. Link Retrieved September 30, 2026.
10. Profound. (2026). Profound raises $180M Series D at $1.8B valuation to build the AI platform for marketing teams, September 15, 2026. Link Retrieved September 30, 2026.
11. Tech.eu. (2026). Conveo raises $50M to scale its AI-powered consumer intelligence platform, September 2, 2026. Link Retrieved September 22, 2026.
12. Branding in Asia. (2026). Bastion Insights introduces AI-powered intelligence platform Bastion Vantage, July 31, 2026. Link Retrieved September 30, 2026.