Key takeaways
- Generative AI is a type of AI that generates human-like text from patterns in large datasets – and it's particularly well suited to qualitative analysis because it can read, summarise, and find themes across large volumes of unstructured content.
- For qualitative researchers, the practical impact is time: tasks that used to take days now take minutes, with time savings of 40–90% depending on the type of research.
- The key to getting useful outputs is prompt quality. Three principles make the difference: start broad, provide context, give clear instructions.
- Indeemo's generative AI tools work across video transcripts, photos, screen recordings, and imported interviews and focus groups – all in a secure environment where your data is never used to train AI models.
- AI handles summarisation, translation, theme detection, and sentiment analysis – so researchers spend more time on the strategic work that actually matters.
What is generative AI, and why does it matter for qualitative research?
Generative AI is a type of artificial intelligence that generates new text based on patterns it has learned from vast amounts of existing data. At its core, it uses large language models (LLMs) – think of GPT as the most widely known example – to understand context, follow instructions, and produce responses that are coherent, relevant, and often hard to distinguish from something a human wrote.
That last part is what makes it useful for qualitative research. Qual has always involved a lot of reading, listening, and pattern-finding across large bodies of unstructured text. Generative AI is built for exactly that kind of work.
At Indeemo, we've been using various forms of AI for several years. The release of ChatGPT in late 2022 accelerated what was already happening – it brought AI awareness into the mainstream and, with it, an explosion of interest in how these tools can change the way researchers work.
One thing worth saying up front: you don't need a technical background to use it well. The term "prompt engineering" tends to intimidate people — it sounds like something invented in Silicon Valley to make a simple idea sound exclusive. There is a craft to writing good prompts, but it's not computer science. It's closer to knowing how to give a clear brief. More on that later.
How does generative AI speed up qualitative data analysis?
Before generative AI, analysing qualitative data meant hours of reading through transcripts, manually coding responses, identifying themes by hand, and pulling together a synthesis that could take days to produce. For video research, it was even more time-consuming – watching footage, taking notes, cross-referencing observations.
Generative AI changes that equation. You give the AI your data – transcripts, video captions, open-ended text responses – and ask it to do the analytical work. It can summarise, identify themes, detect sentiment, and extract quotes in the time it takes to write a few sentences of instruction.
The time savings are real. Based on what we see across projects on the Indeemo platform, researchers typically reduce analysis time by between 40% and 90%, depending on the type of research and what they're asking the AI to do.
What researchers who have run A/B comparisons – analysing data themselves first, then asking the AI to do the same – consistently tell us is that the AI's objectivity sometimes surfaces things they missed. Not because the AI is smarter, but because it doesn't have the same preconceptions about what it expects to find. That's worth knowing.
The broader shift is well documented. A 2024 Qualtrics Market Research Trends Report found that 47% of researchers were already using AI to analyse large qualitative datasets and that share has only grown since.
What types of data can generative AI analyse?
Generative AI on the Indeemo platform works across video transcripts from diary studies, mobile ethnography, and other in-the-moment research; photos with captions and written text entries; screen recordings with voice-over; interviews and focus groups imported from Zoom or Microsoft Teams; and mixed-method bodies of data where in-the-moment mobile uploads and synchronous interview footage are analysed together.
That last one is worth pausing on. Historically, diary study data and interview data lived in separate places and got analysed separately. Indeemo lets you bring them together into a single research repository and look at the full picture as one coherent dataset.
Is my data safe when I use generative AI?
Yes – and this matters, especially for enterprise clients and healthcare research. Indeemo's generative AI integration is built with data security as the starting point. Everything you prompt and every result you get back is owned entirely by you. Your data is not used to train any AI models.
The platform is ISO 27001, SOC 2 Type II, and HIPAA certified. For teams working with sensitive participant data, that's not a footnote – it's a requirement.
What does generative AI make possible that wasn't viable before?
Speed changes what's achievable. When analysis takes days, you run smaller studies. When it takes minutes, you can run bigger ones – and ask more of them.
Some things that were previously impractical are now straightforward. Analysing hundreds of hours of video used to take weeks; it now takes a fraction of that time. That means larger sample sizes, more markets, more data, without a proportional increase in cost.
Mixed-method synthesis is another one. Combining in-the-moment mobile research with focus group or interview footage in a single analysis pass gives a richer picture than either source alone and Indeemo lets you do that in one repository rather than stitching things together across separate tools.
There's also a bigger shift happening in what qual can be used for. Segmentations and brand tracking have traditionally belonged to quant research because the scale required made qualitative approaches unworkable. With the ability to analyse thousands of responses at speed, that's no longer necessarily true.
And then there's the trust question. As AI-generated text becomes harder to distinguish from human-written content, video provides something text alone can't: a real person, in their real environment, captured in the moment. That's part of why we think video research will keep growing, and why the ability to analyse it quickly matters. A 2024 Getty Images study of 30,000+ adults found that 98% of consumers agree authentic images and videos are pivotal in establishing trust – a signal that real, human-captured footage carries weight that synthetic content simply can't replicate.
As Eugene Murphy, Founder and CEO of Indeemo, puts it:
"Generative AI enables researchers to literally have a dialogue with their data. It acts as a research assistant to do tasks such as summarisation and translation leaving researchers with more time for synthesis and strategy."
– Eugene Murphy, Founder & CEO, Indeemo
How do you write prompts that actually work?
This is where most people get stuck – not because prompting is hard, but because the first instinct is usually wrong. When you sit down with a new AI tool and a folder of transcripts, the temptation is to ask it everything at once: analyse all of this and tell me what's interesting. It rarely works well.
Good prompting follows three principles.
Start broad, then narrow
Begin with a wide question to get your bearings – ask the AI for a quick overview of what's in the data, what topics come up, what the general patterns are. This gives you a map before you start navigating.
Then, once you have a sense of the landscape, you can start refining. Ask follow-up questions. Go deeper on specific themes. Request more detail on a particular segment or time period. This iterative approach – starting open and gradually narrowing – consistently produces better outputs than trying to do everything in one go.
Give the AI context
Generative AI doesn't know what your research is about unless you tell it. Before asking it to analyse anything, set the scene: who the participants are, what the research objective is, what you're trying to understand.
There's a balance to strike here. Too little context and the AI produces generic outputs that could apply to any dataset. Too specific too early and it can miss the broader patterns. A good starting prompt might sound like: "These are transcripts from 20 video diary entries. Participants are shoppers aged 25–45 who recently switched supermarkets. The research objective is to understand what drove the switch. Please give me an overview of the main topics that come up across the responses." That's enough to be useful without being so prescriptive that the AI only finds what you've told it to look for.
Give clear instructions
Vague instructions produce vague outputs. If you want themes, say you want themes – and say how many, what level of detail, and whether you want supporting quotes. If you want sentiment, specify whether you want an overall rating or a breakdown by participant group. The more specific the instruction, the more useful the result.
Indeemo's prompt library is organised into categories that map to the most common analytical tasks:
The top prompts – Find the Main Themes, Pull the Best Quotes, Quick Overview, Summary for Leadership, and Top 5 Takeaways – cover the tasks researchers come back to most often.
How does Indeemo's generative AI work in practice?
Indeemo is an end-to-end video research platform, which means generative AI runs through the whole workflow rather than sitting as a separate analysis step at the end.
The process from first participant to final stakeholder presentation: recruit from a global panel of 3 million+ participants in hours; research in 30+ languages using a mobile app that participants take wherever the research is happening – their kitchen, their commute, the shop floor; analyse using AI-powered transcription, translation, theme detection, sentiment analysis, and quote extraction; then create subtitled highlight reels for stakeholders in minutes – short, shareable video clips that bring participants to life and make the insight hard to ignore.
For teams running mixed-methods research, video interviews and focus groups imported from Zoom or Microsoft Teams sit alongside in-the-moment mobile data in the same repository, analysed together as a single body of work.
Do you need to be a research expert to use generative AI for qual?
No. Whether you're an experienced researcher who runs qual studies every week, or a brand team exploring video research for the first time, the platform is built to be accessible. The AI does the analytical work, and the prompt library gives you a starting point so you're not writing instructions from scratch.
If you need more than the platform, Indeemo's Catalyst team can support you at every stage: study design, recruitment, moderation, analysis, or the full project. If you have research ambitions but not the capacity to run them, we can help.
Indeemo can be more than a platform. It can be a partnership.
FAQs about Mobile Ethnography
No. Whether you're an experienced researcher or a brand team exploring video research for the first time, the platform is built to be accessible. If you need more, Indeemo's Catalyst team can support you at every stage: study design, recruitment, moderation, analysis, or the full project.
Based on projects run through the Indeemo platform, researchers typically see time savings of 40–90%, depending on the research type and what they're asking the AI to do.
Yes. Your prompts and results are owned entirely by you — your data is not used to train any AI models. The platform is ISO 27001, SOC 2 Type II, and HIPAA certified.
Video transcripts, photo captions, written text entries, screen recordings, and imported interview and focus group footage from Zoom or Microsoft Teams — including mixed-method datasets.
No. Writing a good prompt is closer to giving a clear brief. Indeemo's built-in prompt library also gives you a starting point for the most common tasks.
Traditional AI interprets and processes existing data. Generative AI generates new text from patterns in vast datasets — meaning it can read transcripts, produce summaries, identify themes, and answer questions about data.
