How Kinsy Researches a Deck: Live Sources, Numbered Citations, and the Fact-Check Pass

By the Kinsy team · Published 25 July 2026

An AI-generated deck you can't trace becomes a liability the moment someone in the room asks where a number came from. This tutorial covers the research side of Kinsy: when live research runs, how numbered citations end up on the slides, what the fact-check pass actually verifies — and a ten-minute routine for checking a cited deck yourself before you present it.

Does Kinsy cite its sources?

Yes — when a topic needs research, Kinsy runs live web research and puts numbered citations on the slides, then fact-checks the claims against the sources before you ever see the deck. The citations are numbered and dated, and they attach to specific claims: the figure, date or finding that came from outside carries its own number, so "where's that from?" has a precise answer rather than a vague wave at a bibliography.

The ordering is the part worth noticing. Research happens before the slides exist, not after. The build runs in front of you, and when a topic calls for research it lands first — you can watch the sources arrive before a single slide is drawn. That means the slides are written from the research; the citations aren't sprinkled on afterwards to decorate claims that were already invented.

When does research kick in?

Kinsy plans every deck's narrative first. Whether it also researches depends on what the deck needs to be right about.

Topics that need current facts get live research. A market overview, a competitor landscape, a lecture that leans on this year's figures — anything where being six months out of date would show — triggers live web research, and the claims that come out of it carry numbered, dated citations.

Argument-led decks may not need it. A talk on why constraints produce better products lives or dies on its reasoning and examples, not on fresh data. Kinsy doesn't pad a deck with citations it doesn't need; research runs when the topic calls for it.

Decks built from your material are grounded in your material. If you attach a document — Kinsy accepts PDF, markdown and plain text, and uploads are free — that file becomes the source of truth, and the fact-check pass verifies the slides' figures, dates, names and quoted findings against it. Live research only adds what your document doesn't contain, and those additions carry citations, so you can see exactly what came from your file and what came from outside it. The full document path is covered in our guide to turning a document into a deck with Kinsy.

What Kinsy citations look like on a slide

A researched deck gives you three things. First, numbered per-claim citations: the number sits on the claim itself, not at the end of the deck. Second, dated sources you can check — each number resolves to a source with a date on it, and the date is doing real work, because a figure's freshness is half of whether you should present it. Third, speaker notes on every slide — what to say, not a restatement of the bullets — so when a cited number draws a question, the surrounding argument is already in your notes rather than in your improvisation.

The fact-check pass is the part most AI tools skip. Generating a citation is easy; verifying that the slide still says what the source says — after the claim has been compressed into slide language — is the hard part, and it's where confident nonsense usually creeps in. Kinsy checks its own claims against the sources before you see the deck, which is also exactly what makes the next section possible.

How to verify a cited deck in 10 minutes

Citations make verification possible; they don't make it automatic. This is practice advice rather than a product feature, and it's worth ten minutes before any deck goes into a room that matters.

1. List the load-bearing claims. Pick the three or four numbers the room will actually challenge — the market size, the growth figure, the date you're hanging the argument on. Ignore the rest for now; nobody interrogates a transition slide.
2. Check each one against its cited source. Follow the number and confirm the source says what the slide says: same figure, same units, same timeframe. Slide language compresses, and compression is where meaning slips.
3. Check the dates. Kinsy's citations are dated for a reason. A figure that was current in January may be folklore by the board meeting — decide for each one whether fresh enough is fresh enough.
4. Rehearse from the speaker notes. Every slide arrives with notes written as what to say. Read them aloud once — if you can't defend a cited claim without the slide in front of you, you don't own it yet.
5. Fix anything you wouldn't defend. Refine conversationally — "replace the market-size figure on slide 6 with our internal number" — or export and change it in PowerPoint. Either way, don't present a claim you've decided not to stand behind.

What citations look like in the exported .pptx

Like everything else Kinsy exports: real text. The numbered markers and source lines get no special treatment because they need none — every text element in a Kinsy export opens in PowerPoint as a real, editable object, never a flattened picture. Restyle the citations to match your house format, reword a source line, or strip them entirely for a teaser deck; it's all ordinary PowerPoint editing. The designed fonts are embedded so the deck survives on any machine, and the file matches the preview you approved to the line break, because the .pptx is native output rather than a conversion bolted on at the end — the reasoning is in our deep dive on why AI deck tools break on export. PDF and per-slide PNG exports are there too, and exporting never costs credits.

Trying the whole loop on a real topic costs nothing:

What research doesn't do

Honesty about the limits, because this is where AI tools usually oversell. Live research grounds a deck in checkable sources; it does not make the deck true by decree, and it cannot know things the web doesn't know. Your Q3 revenue isn't on the internet — internal numbers should come from your own document, uploaded, so the fact-check pass anchors on your file rather than on the public web's best guess about your business.

And a citation makes a claim checkable, not correct. Sources disagree, go stale, and are occasionally just wrong — which is why the citations carry dates, and why the ten-minute check above exists. Research is grounding, not a substitute for your own judgement, least of all on your own numbers. What it removes is the classic AI failure mode: the confident, sourced- from-nowhere figure you only discover on stage. For how the rest of the field handles this, see our guide to AI presentations with citations; for everything either side of the research step — briefs, refinement, presenting — the full tutorial on how to use Kinsy picks up where this one stops.

Keep reading

• The best AI deck maker in 2026: we tested the field
• Can ChatGPT make a PowerPoint? What you actually get
• AI presentation makers that export truly editable PowerPoint files