Write Competitor Battlecards from Public Research

Workflow area: Marketing · Tier: Advanced · Hands-on time: ~45 minutes

The failure mode of AI-written battlecards is confident staleness: models "know" competitors from training data that may be two years old, and they fill gaps with plausible fiction. The discipline this playbook enforces is simple — the model may only synthesize evidence you collected, every claim carries a source and a date, and anything unverifiable is marked as such. A battlecard sales trusts is a curated evidence file wearing a usable format.

When to use — and when not to

• Use it for the 3-5 competitors that appear in real dealsyour CRM's closed-lost reasons tell you which; battlecards for competitors you never actually face are shelf-ware.
• Use it on a refresh cyclequarterly, or triggered by a competitor's launch, pricing change, or funding round. The date on every claim makes staleness visible.
• Use it to arm new sales hiresa sourced battlecard transfers competitive fluency in an hour that otherwise takes two quarters of lost deals.
• Skip claims from private or dubious channelsnothing from an NDA'd deck a prospect showed you, a former employee's confidences, or unverifiable rumor goes on a card your team will repeat in deals.
• Skip trash-talk formatsbattlecards that mock competitors backfire the moment a prospect likes them. Every counter must be respectful and factual; assume the card leaks.

The recipe

1. 1.

   Collect the evidence pack — you, not the model's memory. Per competitor, 20–30 minutes across: their pricing page, two or three product/feature pages, their positioning homepage headline, recent release notes or changelog, a G2/Capterra review skim (sort by recent), any recent funding or launch coverage, and your own CRM's last five win/loss notes against them. Save text snippets with URLs and access dates into one doc. If you use a browsing-enabled model or deep-research mode for collection, treat its output as leads to verify, not evidence — click through before anything enters the pack.

2. 2.

   Run the synthesis pass on your evidence only.

   Copy

   Build a competitor battlecard from the evidence below. Use ONLY
   this evidence — no background knowledge about this company. Every
   claim must cite its source number and date. Where evidence is
   missing or older than 6 months, write [UNVERIFIED] or [STALE —
   date] rather than filling the gap.
   
   Us: [YOUR PRODUCT, ICP, ONE-LINE POSITIONING]
   Competitor: [NAME]
   
   Evidence pack:
   [1] Pricing page, accessed 2026-08-05: "..."
   [2] G2 reviews (last 90 days), accessed 2026-08-05: "..."
   [3] Closed-lost note, deal #482, 2026-07: "..."
   [continue numbering]
   
   Produce these sections:
   1. SNAPSHOT: who they are, who they sell to, their positioning
      claim in their own words [cite]
   2. PRICING SIGNALS: model, published tiers, what's NOT public
   3. WHERE THEY WIN: 3 honest strengths, with evidence — sales
      only trusts a card that admits these
   4. WHERE WE WIN: 3-4 differentiators, each tied to evidence of
      their gap AND our capability
   5. LANDMINES: questions our rep should get the prospect asking
      early, phrased neutrally (e.g. "ask how implementation is
      staffed"), each tied to the weakness it exposes [cite]
   6. OBJECTION COUNTERS: for the 3 objections in our loss notes,
      a 2-3 sentence factual response — acknowledge, reframe,
      evidence
   7. DISQUALIFIERS: deal shapes where they genuinely fit better
      and we should qualify out
   8. SOURCES: numbered list with URLs and access dates
   

3. 3.

   Audit the card against the pack. The model will occasionally launder training-data "knowledge" into a cited section. Check every claim in sections 2, 4, and 5 against its cited source — the load-bearing sections. A trustworthy landmine entry looks like:

   Copy

   LANDMINE 3: "Ask them what's included in the base tier versus
   paid add-ons for reporting." — Their advanced reporting moved to
   a paid add-on in the March repackaging [1]; four recent G2
   reviews cite surprise add-on costs [2]. [Checked 2026-08-05]
   

   If a claim's source doesn't actually support it, cut it. A battlecard with seven bulletproof claims beats one with fifteen shaky ones — reps find out which claims are shaky in front of prospects.

4. 4.

   Pressure-test with a red-team pass. In the same chat: "Now argue the competitor's side: how would their best rep counter each of our WHERE WE WIN points and neutralize each landmine?" Fold the strongest counters back into the card as "if they say X, respond Y" lines. A card that has already survived the counter-argument is what makes a rep confident live.

5. 5.

   Publish in Notion with freshness machinery. One page per competitor from a shared template: the card, an evidence appendix, last_verified date on the page and per-claim, an owner, and a "field intel" comment thread where reps drop what they hear in deals (marked unverified until sourced). Set a quarterly review reminder; the refresh is fast — re-collect the pricing page and recent reviews, ask the model to diff old evidence against new: "What changed, and which claims on the card does each change affect?"

Make it yours

• Loss-note mining first: if you have 20+ closed-lost notes, run a pre-pass — "cluster the stated reasons we lost to [X]" — and let the real loss patterns choose which sections deserve depth.
• Tiered cards: produce a 5-line "glance card" (positioning line, top 2 wins, top landmine, disqualifier) for SDRs from the full card — same evidence, two altitudes.
• Win-story pairing: attach one anonymized customer story per WHERE WE WIN point from your case-study library — a differentiator with a story beats a differentiator with an adjective.

Quality check

• Every claim carries a source and a dateand the section 2/4/5 claims were audited against their sources by a human.
• The card admits where they winreps stop trusting (and using) cards that pretend the competitor has no strengths.
• Landmines are neutral questions, not attackseach would sound fair if read aloud to the competitor.
• The red-team counters are folded inyour differentiators survived the other side's best argument.
• Freshness is visiblelast_verified dates mean a stale card looks stale instead of confidently wrong.