Maptionnaire's analysis tool can generate a written report from your open-ended survey answers. This article explains how that report is built. Below you'll learn how to set it up and how to edit it. If you are interested in understanding more about how it's designed, what's happening behind the scenes and wthe AI-model, keep reading until the end of this article.
Go to the analysis tool. Click the AI symbol and select Generate report automatically.
Click Generate report.
AI generates a report. Go through it and edit it if needed (for texts, click the pen symbol; for charts, just click the chart to access the editing panel).
You can also edit the style of the report.
Finally, download the report in one of the available file formats or copy the link to the online format.
How the AI writes the narrative
Language models are unreliable when you hand them thousands of open answers and ask for a summary. Asked to summarise, a model isn’t counting — it has no dependable way to track how often something was actually said, and what it surfaces depends partly on where the answer sat in the pile.
So Maptionnaire never asks the model to summarise your raw data. Open answers go through three separate rounds instead:
Round 1 — the themes. For each open question, the AI reads the answers and proposes the themes that actually recur in them. It returns at most ten tags per question, and is instructed to prefer fewer.
Round 2 — the tagging. The AI then goes through the answers one at a time and assigns each answer the tags that fit it. An answer can get several tags, or none. Very short answers are skipped.
Round 3 — the narrative. Only now does the AI write. And it does not see your responses at this point: it sees the tag statistics (“mentioned in 87 of 421 answers”) and up to five example answers per theme. A theme that one person raised cannot be written up as a common concern, because the count says otherwise.
The practical consequence: the narrative is a description of the tags. If the tags are right, the narrative is right.
Why it’s built this way
This isn’t a workaround we invented. It is how qualitative researchers have analysed open text for decades — content analysis: agree a set of codes, apply them to every response, then report how often each one occurs. Maptionnaire does the same thing, with the AI as the coder and you as the reviewer.
That split matters, because it plays to what language models are actually good at. Research comparing AI to human coders consistently finds that models are far more reliable at applying a set of categories to responses than at inventing the categories in the first place — and when applying a fixed set, they are at least as consistent as human coders. Round 2 is the reliable part. Round 1 is the step worth your attention, which is why you can replace it with your own categories at any time.
Maptionnaire follows the AAPOR task force’s guidance for AI-assisted coding of open-ended responses. AAPOR — the American Association for Public Opinion Research — published Responsible AI Integration in Survey Research in 2026, and it addresses this exact task: an AI sorting open answers into themes. Point by point:
| The guidance | How Maptionnaire does it |
|---|---|
| Keep labelling consistent by working from a fixed codebook | Rounds 1 and 2: settle the tags first, then apply that same set to every answer |
| Have people review and correct the AI’s codes before analysis | Every tag is visible next to the answer it was applied to, and you can change it |
| Keep the reasoning traceable | Each claim traces back: narrative → tags → the individual answers, all open to inspection |
| Be explicit about which decisions are the machine’s and which are yours | See “What the AI does not decide” below |
| Use AI to support the analyst, not to replace them | The report is a first draft: every section can be rewritten, reordered or removed |
If you’re worried about bias, write the themes yourself
The clearest answer to “how do I know the AI didn’t slant this?” is not to take our word for it. Click Define tags, type the themes you want, and re-tag. The AI is then sorting answers into your categories, and its own reading of the data no longer shapes what the report can say. Everything downstream — the counts, the narrative, the quotes — follows from tags you chose.
This is also the more rigorous way to work if you already have a coding frame, a policy vocabulary, or themes from an earlier round of engagement that you need this survey to be comparable with.
Checking the tags yourself
You don’t have to take the tagging on trust, every tag the AI assigned is visible and editable in the analysis tool.
- Add the open question to your analysis view. The answers appear as a table, with each answer’s tags next to it.
2. Read ten or so answers and check the tags look sensible. Use the tag filter above the table to look at one theme at a time — that is usually the fastest way to spot a tag that is being applied too loosely.
3. Open the Tags tab to see how many answers carry each tag.
Worth a closer look if your survey ran in a less widely spoken language: AI models are trained on far more English than Finnish, Somali or Greenlandic, and the tagging tends to be correspondingly less sure-footed.
If something is off, you can fix it at either level:
- A single answer is tagged wrong: edit its tags directly in the table (or in the response popup on the map). Hand-applied tags behave exactly like AI-applied ones: they filter the map, count in the tag chart, and come along in the Excel export.
- The whole set of tags is wrong: click Define tags, type the themes you want the answers grouped into, and re-tag. The AI then sorts the answers into your tags instead of its own.
If you change tags after generating a report, Maptionnaire tells you which sections now describe tags that no longer hold, and offers to refresh them.
In the report itself, click a section and choose Behind this section to see which tags that section was built from.
Quotes are copied, not written
The AI does not type the quotes in your report. It picks an example answer by number, and Maptionnaire inserts that respondent’s text verbatim. A quote in the report is therefore always something a respondent actually wrote — it cannot be paraphrased, merged or invented.
Cluster analysis is arithmetic, not AI
The map cluster sections work the same way: the analysis is mechanical, and only the wording is written by AI.
- Clusters are found with a standard density-based clustering algorithm (DBSCAN), which groups map pins that sit close together. You control it: From Clustering settings, set the cluster radius and density, preview the clusters on a map, and pick which ones the report should cover
- For each cluster, Maptionnaire compares the responses inside it against the responses in the rest of the survey, and keeps the differences that are statistically significant. You can open the list of these differences per cluster and read the numbers.
- The AI’s only job is to put that comparison into a sentence.
What the AI does not decide
Charts, chart types, figures and percentages are computed from your response data in the same way as everywhere else in the analysis tool. The AI groups elements into sections and writes prose; it never produces a number.
And nothing is final: every section can be edited, reordered or removed, and the underlying views are ordinary saved analysis views you can open and change.
The model, and where your data goes
As of August 2026, Maptionnaire uses Claude Sonnet 4.6, running on Amazon Bedrock — the same AWS infrastructure that hosts the rest of Maptionnaire. Every step described above uses it: reading the answers, tagging them, and writing the narrative. Your responses are processed in the same part of the world where they are already stored. The model does not retain your responses after processing them, and does not use them to train models.