How to Cut Debate Cards: The Complete Guide (2026)

By Zaid Anwar, Founder & Developer, DebateCardAI · Published April 1, 2026 in Guides

Cutting a card means taking a passage from a real source and formatting it so you can read it in a round and defend it when someone challenges it. Everything else is detail.

Most guides describe that process. This one runs a single card through it, start to finish, using a real report you can go open right now. Each step shows you what it produces.

What a Finished Card Looks Like

Every card has three parts: a tag (your one-sentence claim), a citation (who said it and where), and the body text (their words, not yours). Here is a complete one, formatted the way it would come out of Verbatim.

A card has four layers, and people mix them up constantly:

• Highlighted is what comes out of your mouth. Only this. Nothing else gets said out loud.
• Underlined is the relevant context: the part of the argument you want visible when a judge or opponent reads the card, but that you do not spend speech time saying.
• Bold is what you emphasize inside what you say. Usually the number.
• Shrunk to 8pt is everything else. It stays in the card so nobody can accuse you of pulling a number out of nowhere, and you never read a word of it.

The highlight is always underlined too, since it is the most relevant context by definition. That is why underlining alone does not mean "read this."

Read only the yellow and you get:

The global energy sector added nearly 2.5 million jobs in 2023 · Employment in energy rose by 3.8%, outpacing the economy-wide average of 2.2% · jobs grew rapidly in both clean energy sectors (1.5 million) and fossil fuels (940 000)

Thirty-nine words that still hang together as speech. That is the test, and it is why the highlight skips over "on the back of rising investment" and lands again on the next number. You are not highlighting sentences, you are highlighting the shortest path through them.

Now look at what got shrunk: the definition of what counts as "energy," and the China and Middle East sentence. Both are true, and neither helps you. Deciding what drops to 8pt is the whole skill.

Step 1: Find a Source Worth Cutting

Judges weigh evidence quality, and good opponents attack weak sourcing before they attack your argument. The rough hierarchy every circuit uses:

The card above is top of that list: an intergovernmental agency publishing its own dataset. That is why the citation can be short and still carry weight.

One rule with no exceptions: quote, never paraphrase. Direct quotation is what lets a judge check you, and the NSDA's evidence rules are built around producible text. If you cannot produce the original source when challenged, rule 7.2.B treats your evidence as non-existent.

Step 2: Write the Tag

Your tag is the first thing the judge hears, so it has to state an argument, not a topic.

Weak: Clean energy and employment Strong: Clean energy added 1.5 million jobs in 2023, and energy employment grew faster than the economy as a whole

The weak version makes the judge work out what you are claiming. The strong version tells them, and it commits you to something your evidence actually proves.

That last part matters more than it sounds. Writing a tag that overstates your source is called power-tagging. Here is the same IEA card with a tag that goes too far:

Power-tagged: Fossil fuel jobs are collapsing as clean energy takes over

Go back and read the body. It says fossil fuel employment grew by 940,000 in 2023. The tag claims the opposite of the evidence sitting underneath it. Any opponent who reads your card catches this, and now the round is about your honesty instead of your argument.

Power-tagging is not itself an NSDA rule violation, since the text was quoted accurately. It is a community norm, enforced by judges on the ballot. The fix is boring: if the source says "may contribute to," your tag cannot say "proves."

Step 3: Build the Citation

A citation has two halves that do different jobs. Look at the card again:

IEA 24 is the spoken half. Rule 7.1.B only requires the author's last name and the year when you read the card, so that is all you say.

Everything in the brackets is the written half, and rule 7.1.C of the NSDA Unified Manual says what belongs there, "to the extent provided by the original source." In order, that is:

The whole written half follows one pattern, and it is worth memorizing because every card you ever cut uses it:

Lastname YY, [Full name, M/D/YY, Qualifications, Title, Publication, DOA: M/D/YYYY URL]

What is that //za2 at the end?

That is a card signature, and it goes immediately after the closing bracket. It is a short tag that marks who cut the card, so that when the card gets shared around, dumped in a wiki, or lifted into someone else's file, the credit travels with it. Mine is //za2.

Pick something short and unmistakably yours, usually initials plus a number, and put the same one on every card you cut. It costs nothing and it is the only reason anyone knows a card in circulation came from you.

What separates a good cite from a compliant one

Hitting all seven elements makes a citation legal. Three things make it actually work in a round:

Qualifications that match the claim. "Intergovernmental energy body, standard source for global energy statistics" is doing real work, because the card is about global energy statistics. A Nobel laureate in literature has impressive qualifications and none of them are relevant to energy employment. Write the qualification that makes this author the right person to answer this question, not the most flattering one you can find.

A date the judge can compare. Evidence gets beaten by newer evidence. Putting 11/13/24 in the cite lets you say "their card is from 2019, ours is from 2024" and have the judge see it. Burying the date costs you that argument.

A URL that actually resolves to the text you quoted. Not the publisher's homepage, not a paywalled abstract, not a search result. If someone calls for your card, rule 7.2.B treats evidence you cannot produce as non-existent, and that is a disqualification. Link the page the words are on.

If you cannot find any qualifications for an author, that is usually the source telling you to find a different one.

Step 4: Highlight the Body

Keep the full paragraph in the card, underline the context that matters, and highlight only the warrant. The most common mistake is highlighting too much, which defeats the point: if everything is emphasized, nothing is.

Here is the same source, highlighted badly:

Count the yellow in each: 110 words here against 39 in the card at the top of this guide, from the same source, proving the same thing. That is nearly three times the speech time for zero extra argument.

The solid block of yellow is the tell. It means nobody decided what the evidence was for. A real card is mostly small grey text, because most of any paragraph is setup, and setup is what your opponent reads, not what you say.

Ask one question of every sentence: does this change what the judge believes? If not, it stays in the card and out of your mouth.

Do Not Get Yourself Disqualified

The NSDA splits evidence violations into ones a judge weighs against you and ones that end your tournament. The short version: distorting a quote or citing evidence you cannot produce is a disqualification under rules 7.4.C and 7.4.D. Clipping, which means claiming you read the whole highlighted portion when you skipped part of it, is an automatic loss with zero speaker points.

Clipping is the one people fall into by accident. In the example card, if you skip "outpacing the economy-wide average of 2.2%" while reading but never mark it, you clipped. Skipping is allowed. Skipping silently is not. Mark where you start and stop.

We wrote a full breakdown of all four violations, what each one costs, and how a challenge actually plays out: what counts as an evidence violation.

Prep Faster

Three things that actually move the needle, past the obvious advice to start early:

Search for meta-analyses. One paper summarizing thirty studies gives you more cards than three individual papers. Add "systematic review" or "meta-analysis" to a Google Scholar query.

Cut in batches. Find every source first, open them all, then format every card in one pass. Switching between researching and formatting is what makes prep feel endless.

Check the paywall before you write the tag. Nothing wastes time like building a citation for an article whose full text you cannot reach. Google Scholar's "All versions" link usually turns up a free PDF.

Doing This With AI

The four steps above are the job. Steps 3 and 4 are mechanical, and that is where DebateCardAI comes in: you paste a source URL and your claim, and it finds the passage, writes the tag, pulls the citation, and highlights the body in the same structure you just read. Evidence Finder handles step 1, searching with debate-specific filters and flagging the passage worth cutting in each result.

What it does not do is decide what argument to run or whether a card is worth reading. Those are still yours, and they are the part that wins rounds.

If you are new, go cut three cards on your current resolution using the format above, and read them out loud to check your highlighting. If you already cut cards and just want it to take less time, try it free: five cards a day, no credit card. For how to use AI in debate without hollowing out your own skills, see our guide on AI in debate.

Written by Zaid Anwar, founder of DebateCardAI and competitive PF debater.