A competitor just won a deal you thought was yours. The rep fumbled when the prospect asked “How do you compare to them?” — because the last battlecard was written six months ago, runs eight pages, and nobody has read it since the all-hands meeting where it was announced. The problem isn’t the rep. It’s the battlecard.

The goal of this guide is to show you how to create a sales battlecard using AI — not as a research summary document, but as a live sales weapon. There is a fundamental difference between a product that lists features and a battlecard that gives reps a three-sentence “kill phrase” for a cold call or a “Landmine Question” that makes the prospect realise the competitor’s weakness on their own. This article teaches you the second approach.

The reason most AI-generated competitive intelligence fails is not the AI — it is the prompt. Asking ChatGPT “What are the weaknesses of Competitor X?” produces a diplomatic, hedge-everything paragraph pulled from stale training data. That is worse than useless in a live sales conversation. What works is feeding the AI real source material — G2 reviews, current pricing pages, earnings transcripts — and then instructing it to output sales-ready scripts, not research summaries. That shift from research to weaponization is the entire methodology this article is built on.

You will also find a clear explanation of which AI tool to use for each task, how to prevent the hallucination problem that makes reps look foolish on calls, and how to protect your own internal positioning data from being exposed to public AI models.

Why Static Battlecards Don’t Close Deals

A sales battlecard is a one-page reference document that gives a sales rep everything they need to win a deal where a competitor is in the room: the competitor’s key weaknesses, the questions to ask that expose those weaknesses without sounding aggressive, the objection rebuttals to use when the prospect pushes back, and the two or three positioning statements that make your product the obvious next step. That is what a battlecard is supposed to be. What most battlecards actually are is a six-page PDF with bullet points from the competitor’s About page.

The first problem is research time. Building a real competitive battlecard manually requires reading the competitor’s entire website, combing through G2 and Capterra reviews, monitoring their pricing changes, and synthesizing all of it into actionable sales language. That takes days. Most teams do it once a quarter and declare the result “good enough” — which it isn’t, because competitor pricing pages change, new features launch, and the reviews keep coming.

Why “Ask ChatGPT About My Competitor” is Dangerous

A common mistake is opening ChatGPT and typing “What are the weaknesses of [Competitor Name]?” This feels productive but is actively dangerous in a sales context. The AI will produce a plausible-sounding paragraph based on training data that may be 18 months old. If a competitor revamped their onboarding process or dropped a problematic pricing model six months ago, ChatGPT may still report the old weakness as current — and when your rep cites it on a call, the prospect (who just finished onboarding with the competitor last month) will know immediately that your intelligence is stale.

The correct approach is zero-shot grounding: you bring the current data to the AI, rather than asking the AI to retrieve it from memory. Every prompt in this article includes an instruction that forces the AI to work only from the material you provide, and to flag any gaps rather than filling them with inference.

The Shift from Research to Weaponization

Here is what actually matters: the purpose of a battlecard is not to inform your sales team — it is to arm them. There is a difference. Information tells reps that a competitor has a long implementation timeline. Weaponization gives them the exact question to ask a prospect in a discovery call that makes the prospect conclude that six-month implementation is a risk they cannot afford — without the rep ever mentioning the competitor by name. That question is called a Landmine Question, and you can generate them in under a minute with the right AI prompt.

The 3 Golden Rules of AI Competitive Intelligence

Before running any prompt, these three principles determine whether your output is reliable enough to hand to a sales rep. Breaking any of them is how teams end up with battlecards that embarrass reps on calls.

With these three rules in place, here are the data sources you should gather before running the prompts below. You do not need all five for every battlecard — pick the ones available for your specific competitor:

• G2 or Capterra 1-star/2-star reviews: Export or copy-paste the most recent 50–100 negative reviews. These are the most valuable raw material in this entire workflow — they are unfiltered buyer complaints.
• Current pricing and features page: Copy the text directly from the competitor’s pricing page right now. Do not rely on a cached or screenshot version from last quarter.
• Recent earnings call transcript or 10-K: For publicly traded competitors, this reveals self-reported strategic vulnerabilities — the ones executives disclosed because they were legally required to.
• Competitor’s job postings: A company hiring heavily for customer success roles is signalling a churn problem. A company with 20 open engineering roles in one product area is scaling something — or fixing something broken.
• Recent press releases or news: Funding rounds, leadership changes, product discontinuations, and pricing restructures all represent potential battlecard angles.

Create a Sales Battlecard Using AI: 5 Copy-Paste Prompts

Each prompt below is structured to produce something a rep can actually use in a meeting — not a research report they will skim once and never open again. Fill in the bracketed sections with your actual product and competitor details, paste your source data below the prompt, and run it in the AI tool specified for that task.

Prompt 1: The G2 Review Sentiment Extractor

This is the highest-value prompt in this entire guide. G2 and Capterra negative reviews are the most honest source of competitive intelligence available — they are real buyer complaints, in real buyers’ own words, about what they hate about the competitor’s product. Reading 100 of them manually takes three hours. The AI processes them in 90 seconds and quantifies the patterns.

To use this prompt: go to G2 or Capterra, filter the competitor’s reviews by 1-star and 2-star, and copy the text of the most recent 50–100 reviews into a document. Paste that document below the prompt. Claude 3.5 is the recommended tool here because of its ability to handle large volumes of text and find patterns across them.

Act as an expert Product Marketing Manager. I have pasted 1-star and 2-star reviews for our competitor, [Competitor Name], from G2/Capterra. Base your entire analysis ONLY on the reviews provided below — do not use external knowledge about this competitor.

MY PRODUCT: [One sentence description of what you sell and who you serve]

TASK 1 — TOP 3 SYSTEMIC FAILURES
Identify the top 3 most common complaint categories across all reviews. For each:
— Category name (e.g., "Customer Support," "Onboarding Complexity")
— Estimated % of reviews that mention this issue
— One direct, anonymized quote from a reviewer that best illustrates the problem
— One sentence on how our product specifically avoids or solves this failure

TASK 2 — BUYER SENTIMENT SUMMARY TABLE
Create a Markdown table with columns: Failure Category | Frequency | Buyer Quote | Our Advantage

TASK 3 — ONE KILL PHRASE
Based on the #1 failure, write a single punchy sentence a sales rep can say on a call that positions us as the solution without naming the competitor directly. Max 20 words.

Output as a formatted Markdown report.

[PASTE G2/CAPTERRA REVIEWS BELOW THIS LINE]

Prompt 2: The Landmine Question Generator

A Landmine Question is a consultative discovery question that makes a prospect realise a competitor’s weakness on their own — without the rep explicitly mentioning it. This is significantly more effective than direct trash-talking, which buyers find off-putting and which damages your credibility. According to research on how B2B buyers evaluate vendors, prospects trust their own conclusions far more than vendor claims. A well-crafted Landmine Question leads the prospect to the conclusion you want them to reach.

To generate these, you need to know the competitor’s most exploitable weakness first (which Prompt 1 will give you). Then feed that weakness into this prompt.

Act as an elite Sales Enablement Coach specializing in consultative selling. I need to generate "Landmine Questions" for my sales team.

A Landmine Question is a polite, curiosity-driven discovery question that leads a prospect to realize a competitor's weakness on their own. It never mentions the competitor by name. It never sounds like a sales tactic. It sounds like a smart, experienced consultant asking about their priorities.

MY PRODUCT: [One sentence description]
COMPETITOR WEAKNESS TO EXPLOIT: [e.g., "Their implementation requires 6+ months and dedicated technical resources"]
PROSPECT ROLE: [e.g., "VP of Operations at a 200-person manufacturing company"]

Generate 5 Landmine Questions. For each question, provide:
1. The question itself (phrased naturally, conversationally)
2. The competitor weakness it is targeting (clearly labelled)
3. The ideal follow-up response if the prospect confirms the concern

Format as a numbered list. Questions must be 1-2 sentences maximum — they should feel like natural conversation, not an interrogation.

Do not write anything that sounds like a sales pitch. The goal is to make the prospect feel smart for identifying the risk, not to make the rep sound clever.

From Weakness to Weapon: See the Difference

Prompt 3: The Live Website Capability Matrix

Use this with Perplexity AI or Microsoft Copilot — both can browse the live web. This prompt generates a feature-by-feature comparison based on what each company currently advertises on their public pages, with explicit instructions to flag anything that is not confirmed rather than guessing. The output is a table your sales team can reference instantly when a prospect asks “How do you compare to them?”

Use your live web browsing capability to review the current features and pricing pages of both companies listed below. Base your comparison ONLY on what is publicly stated on each site right now — do not use training data or general knowledge about either product.

MY COMPANY'S FEATURES PAGE: [Your URL]
COMPETITOR'S FEATURES PAGE: [Competitor URL]

TASK 1 — FEATURE COMPARISON TABLE
Create a Markdown table with 4 columns:
Feature / Capability | Our Product | Competitor | Verdict (We Win / They Win / Parity)

RULES:
— Only include features where there is a clear difference. Skip features where both products are equivalent.
— If a feature is not explicitly mentioned on a company's page, write "NOT ADVERTISED" — do not assume it exists.
— Do not hallucinate. If you cannot access a URL, say so immediately and stop.

TASK 2 — TOP 3 FEATURES WHERE WE WIN
For each, write a 2-sentence sales-ready explanation a rep can use when a prospect asks "What makes you better?"

TASK 3 — TOP 2 FEATURES WHERE THEY WIN
For each, write a 2-sentence honest acknowledgement plus a reframe ("This matters less if your priority is X, where we...").

Do not use the word "synergy," "innovative," or "game-changing."

Prompt 4: The Instant Objection Rebuttal (“Why Should I Switch?”)

This is the prompt to build the most-used section of any battlecard: the objection handling scripts. Every prospect who is already using a competitor will eventually say “We’re happy with what we have” or “Why would I switch?” This prompt generates a three-part response that acknowledges their loyalty without dismissing it, introduces the specific pain point their current vendor is creating, and positions your product as the natural next step — all without being pushy.

Act as a senior B2B sales trainer. I need to build an objection handling script for my sales team for when prospects say "We already use [Competitor Name], and we're happy with it."

MY PRODUCT: [One sentence — specific, not generic]
COMPETITOR NAME: [Name]
COMPETITOR'S KNOWN WEAKNESS (from our G2 research): [Insert the top weakness from Prompt 1]
PROSPECT'S LIKELY ROLE: [e.g., "IT Director at a 300-person financial services firm"]

Write a 3-part objection rebuttal script:

PART 1 — ACKNOWLEDGE (1 sentence)
Genuinely acknowledge that the competitor is a legitimate choice and that many good teams use them.

PART 2 — THE PIVOT (2 sentences)
Introduce one specific pain point that companies at their stage often don't notice until it creates a real problem — tied to the known competitor weakness. Do NOT start with "However" or "But."

PART 3 — THE SOFT NEXT STEP (1 sentence)
End with a low-friction question that opens a conversation rather than asking for a commitment.

Then write a "Quick Version" — the same rebuttal in 3 sentences or fewer that an SDR can memorize and use on a cold call.

Tone: confident, peer-level, not sales-y. Do not use: "just," "quick chat," "circle back," "touch base," "synergy."

Prompt 5: The 10-K Strategic Vulnerability Finder

Use this for enterprise-level deals against publicly traded competitors. Every publicly traded company files quarterly earnings calls and annual 10-K reports with the SEC — and in these documents, their executives are legally required to disclose material business risks. This means that when a CEO says “we’re seeing elevated churn in our mid-market segment” or “our EMEA expansion has faced delays,” those are verified, disclosed vulnerabilities you can build positioning against.

Download the competitor’s most recent 10-K from the SEC EDGAR database or find their earnings call transcript online, and upload it to Claude (which handles large documents most reliably). This prompt extracts the CEO’s own admissions and converts them into sales angles.

Act as a Sales Director preparing for enterprise-level competitive conversations. I have attached a document — either a quarterly earnings call transcript or a 10-K annual report — for our competitor, [Competitor Name]. Base your analysis ONLY on the provided document — do not use external knowledge.

MY PRODUCT: [One sentence — what you sell and who your primary buyer is]

TASK 1 — STRATEGIC VULNERABILITIES (executive-disclosed)
Read the QA section of the earnings call or the Risk Factors section of the 10-K. Identify 2–3 areas where executives admitted to:
— Elevated churn or customer loss in a specific segment
— Product gaps or delayed feature development
— Geographic market weakness or slow expansion
— Competitive pressure from specific categories of competitor

For each vulnerability:
— Quote the relevant passage (under 25 words, directly from the document)
— Explain in one sentence how our sales team can reference this in an enterprise conversation

TASK 2 — ENTERPRISE POSITIONING ANGLE (1 paragraph)
Write a brief that a Sales Director can use to frame the competitive conversation at a CFO or VP-level meeting. The tone should be strategic, not aggressive — we are framing this as reducing risk for the buyer, not attacking the competitor.

TASK 3 — ONE BOARDROOM-READY QUESTION
Write a single, sophisticated question a Sales Director can ask a C-suite prospect that subtly draws their attention to the competitor's disclosed vulnerability — without reading like a sales trap.

[PASTE EARNINGS TRANSCRIPT OR 10-K EXCERPTS BELOW THIS LINE]

Perplexity vs. Claude vs. ChatGPT: The Best AI for Competitive Research

The practical recommendation in one sentence: use Perplexity for anything requiring live web data, use Claude for anything requiring large document analysis, and use ChatGPT for generating conversational sales scripts. Here is the full breakdown.

Why Perplexity AI is the Standard for Live Audits

Perplexity AI is built specifically around live web research with cited sources. When you ask Perplexity to audit a competitor’s current pricing page or pull their most recent product announcement, it browses the live web and cites the source URL for each claim. This citation feature is critical for competitive intelligence — you can verify each fact before it goes into a battlecard, which prevents the hallucination problem entirely. Microsoft Copilot with Bing integration is a strong alternative for teams already in the Microsoft 365 ecosystem, and it has the advantage of being usable within your organisation’s security boundary.

Why Claude Wins for Analysing G2 Reviews and Large Documents

When you have 100 G2 reviews to process or a 60-page 10-K to analyse, Claude 3.5 Sonnet is currently the strongest choice. Its context window can hold substantially more text than most other tools, which means it can find patterns across an entire review dataset without truncating the input. It also tends to produce more analytically rigorous output for synthesis tasks — identifying which complaints appear across multiple reviews rather than just summarising the most recent ones.

CRITICAL: Will the AI Tell My Competitor I’m Researching Them?

This question comes up in almost every conversation about AI competitive intelligence, and the short answer is no — AI tools do not alert the company you are researching. You are asking an AI to help you analyse publicly available information. That is not meaningfully different from using Google to research a competitor, except that you are doing it much faster.

The real privacy concern runs in the other direction: will your own sensitive competitive positioning data be exposed through the AI tool you are using? That is a legitimate risk if you are pasting your internal battlecards, your product roadmap, or your proprietary go-to-market strategy into the free tier of a public AI tool. Here is how to manage that risk:

Data Safety Framework for Competitive Intelligence Work

The legal question — is it legal to scrape a competitor’s public website for analysis? — is generally yes for publicly available content that you are reading manually or having an AI summarise from. You are not accessing anything behind a login or using automated tools that violate a site’s terms of service. You are reading their public marketing page and asking an AI to summarise it. For any concerns specific to your industry or jurisdiction, consult your legal team. For teams using Microsoft 365 and wanting to keep all of this work inside a secure environment, our guide to Microsoft Copilot for enterprise use covers the full capability and security picture. Teams managing their battlecard data in Excel before feeding it to AI will find Copilot in Excel useful for that prep step.

Frequently Asked Questions

Your Next Steps

• 1

  Pick your top two competitors and run Prompt 1 this week

  Go to G2 or Capterra, filter each competitor’s reviews by 1-star and 2-star, and copy the most recent 50 reviews into a document. Run Prompt 1 in Claude. The output — a quantified failure table with direct buyer quotes — is the foundation every other prompt builds on. Do this before anything else.

• 2

  Generate Landmine Questions from the top weakness you found

  Take the number one failure category from your Prompt 1 output and feed it into Prompt 2. Run it in ChatGPT Plus. Share the five generated Landmine Questions with two or three reps and ask them to test them in their next discovery calls. Collect feedback on which ones land most naturally.

• 3

  Build the one-page format using Copilot in Word

  Take the outputs from Prompts 1, 2, 3, and 4 and consolidate them into a single document. The layout should be: Competitor Overview (2 sentences), Their Top 3 Weaknesses (from Prompt 1), Our Top 3 Win Positions (from Prompt 3), Five Landmine Questions (from Prompt 2), Objection Rebuttals (from Prompt 4), Kill Phrase (from Prompt 1). Keep it to one page. Copilot in Word can help format and tighten the language once the content is assembled.

• 4

  Connect the battlecard to your post-deal learning loop

  A battlecard built from G2 reviews is a starting point — the best competitive intelligence comes from your own lost deals. After every deal lost to a specific competitor, feed the call transcript and loss notes into your AI win/loss analysis workflow and use the findings to update the relevant battlecard section. Over time, your battlecard becomes a living document grounded in real buyer conversations rather than public review sites alone.