Purpose: These are the most common failures when AI applies cognitive biases to marketing copy. Each anti-pattern describes what the AI produces, why it fails psychologically, and the concrete fix. Check your output against this list before delivering.
What AI writes:
"Thousands of satisfied customers trust our product."
Why it fails: The brain treats vague numbers as noise. "Thousands" could mean 2,001 or 99,999. No specificity = no credibility. Social proof without a number, face, or name is indistinguishable from fiction.
Fix:
"4,827 teams switched to [Product] last month. Here's what Sarah Chen, VP of Ops at Merge, told us: 'We cut reporting time from 11 hours to 17 minutes.'"
Rule: Social proof needs at least TWO of: specific number, full name, role, photo, measurable result.
What AI writes:
"Every day you wait, your competitors are stealing your market share. Your business is bleeding revenue. Act now."
Why it fails: Fear activates the amygdala - fight/flight/freeze. If you trigger fear but don't immediately provide the safe path, the brain goes to "freeze" (scroll past) or "flight" (close the tab). Fear without resolution = anxiety without action.
Fix:
"Competitors who adopted [solution] grew 34% faster last year. The gap compounds monthly. Here's the 3-step switch that takes 15 minutes and costs nothing to try: [Step 1 → Step 2 → Step 3]."
Rule: Every fear trigger must be followed within the SAME paragraph by a concrete, low-effort solution. Fear opens the wound. Solution closes it. Never leave fear hanging.
What AI writes:
"This isn't just a CRM. It's a revolution in how teams collaborate."
Why it fails: Framing tells the brain "this is different." But without an anchor (a reference point), "different" has no scale. The claim floats in space. The brain discards unsupported claims.
Fix:
"This isn't a CRM. Traditional CRMs take 6 weeks to set up and cost $50/user/month. [Product] connects to your existing tools in 4 minutes. $19/user. No setup. No consultants."
Rule: Every frame needs an anchor. "Not X, but Y" only works when X is concrete and quantified. Anchor = the old way, stated with specifics. Frame = your way, contrasted against it.
What AI writes:
"Studies show that our method is the most effective."
Why it fails: "Studies show" is the most overused, credibility-destroying phrase in marketing. It signals "I read a headline once." The brain has been burned by this too many times.
Fix:
"A 2023 meta-analysis in the Journal of Applied Psychology (N = 12,000) found that [specific finding]. We built [Product] to operationalize that exact mechanism."
Rule: Authority claims need: named source, year, specific finding. "Harvard research" = weak. "Harvard Business Review, March 2023: 'Teams using async communication ship 23% faster'" = strong.
What AI writes:
"Here's a free ebook! Now buy our course."
Why it fails: Reciprocity works when the gift feels genuine, not when it's obviously a hook. If the "free value" is a thinly veiled sales pitch, it triggers reactance - the brain's "you're trying to manipulate me" alarm.
Fix: Give genuine standalone value FIRST. The ebook/article/tool should be valuable even if the reader never buys anything. The pitch comes in a SEPARATE communication, after the value has been consumed.
Rule: The gift and the ask must live in separate pieces of content. Email #1 = pure value. Email #2 = social proof. Email #3 = soft pitch. The separation IS the reciprocity mechanism.
What AI writes:
"Only 3 spots left! Offer expires in 2 hours! ⏰"
Why it fails: Consumers have been burned by fake countdown timers and artificial scarcity for decades. The brain has developed skepticism. If your urgency mechanism resets on reload, is never explained, or feels arbitrary - trust evaporates instantly and permanently.
Fix (genuine scarcity):
"We're capping this cohort at 50 founders. Not because we want to - because our team of 3 can't give personalized feedback to more than 50. Current count: 47. Registration closes Friday at midnight, period."
Rule: Scarcity must be EXPLAINED, not just DECLARED. Why is it limited? Capacity? Time? Attention? Cost? If you can't explain the constraint, don't use scarcity.
What AI writes:
"You've come this far. Don't stop now." (in the FIRST email or FIRST paragraph)
Why it fails: Sunk cost only works after the prospect has ACTUALLY invested something: time, attention, effort, emotion. Invoking it too early is absurd - "I just opened this email, what 'far' have I come?"
Fix: Build investment first across multiple touches. THEN invoke sunk cost:
"You've watched the 90-minute webinar. You've read the 12-page guide. You've seen the case studies. The only thing between you and the outcome is one decision."
Rule: Sunk cost is a LATE-STAGE bias. Use only after the prospect has demonstrably invested attention. Never in first contact.
What AI writes:
"Join our community of like-minded people."
Why it fails: "Like-minded" is empty. It means nothing. In-group favoritism requires a SPECIFIC shared identity. The brain can't feel belonging to "people who are like-minded about unspecified things."
Fix:
"Join 3,200+ engineering leaders who've left Big Tech to build independent businesses. We share war stories, client leads, and the actual revenue numbers nobody posts on LinkedIn."
Rule: In-group requires: named identity (who), shared experience (what you've been through), and an out-group (who you're NOT). "Engineering leaders" (in-group). "Big Tech" (out-group). "Left to build independently" (shared experience).
What AI writes:
"As seen on CNN, Forbes, and TechCrunch!"
Why it fails: The halo only transfers if the audience RESPECTS the source. A DTC audience may not care about Forbes. A developer audience may actively distrust TechCrunch. The wrong halo source can create negative transfer.
Fix: Match the authority source to the audience's actual trust network. Developers trust GitHub stars and specific engineers. Designers trust Dribbble and specific designers. Founders trust YC, Indie Hackers, specific founders they follow.
Rule: Don't guess which authorities matter. ASK in Deep Mode. Or default to peer-level proof (people like them) rather than institutional proof (publications).
What AI writes:
"Most people are wrong about [topic]. Here's the truth."
Why it fails: This reads as "you're probably one of the stupid people." Even if the reader agrees with the take, the framing makes them defensive. Nobody wants to be told they're in the "most people" who are wrong.
Fix:
"You've probably sensed that [common belief] doesn't add up. You're right. Here's why - and what the data actually shows."
Rule: Confirmation bias marketing must make the reader feel SMART for doubting, not DUMB for believing. "You were right to question this" > "Everyone else is wrong."
What AI writes:
"Imagine you're struggling with [problem]."
Why it fails: "Imagine" + generic scenario = nothing. The brain can't visualize "struggling with productivity." It CAN visualize "staring at a blinking cursor at 2 AM with a deadline in 6 hours."
Fix: Supply the concrete, sensory details. Smells, sounds, specific objects, exact times, real names.
"It's 2:14 AM. Your deadline is 8 AM. You've written 47 words. The cursor blinks. Blinks. Blinks. You open Twitter. Close it. Open it again. You've been 'working' for 6 hours and have nothing. That feeling? That's what we fix."
Rule: A story is only available to memory if it contains at least 3 concrete, sensory details. Abstract scenarios are invisible to the brain's recall system.
What AI writes:
"You say health is your priority, but you haven't exercised in months. Why?"
Why it fails: Calling out the reader's hypocrisy triggers shame, not action. The brain's response to shame is withdrawal, not engagement. You've created dissonance, but the resolution path is "I'm a bad person" - and nobody clicks "buy" from that state.
Fix:
"Most of my clients tell me health is their #1 priority. Then I ask when their last checkup was. Silence. Not because they don't care - because the system makes it hard. Appointments during work hours. Forms that take 30 minutes. Waiting rooms. We fixed that."
Rule: Dissonance must be resolved by blaming the SITUATION, never the PERSON. "The system failed you" > "You failed."
Before delivering any output, scan for these failures:
- Social Proof: Does it have a specific number + name/role/face?
- Fear: Is the solution in the SAME paragraph as the threat?
- Framing: Is the "old way" anchored with specifics?
- Authority: Is the source named + dated + finding quoted?
- Reciprocity: Is the value separable from the pitch?
- Scarcity: Is the constraint EXPLAINED, not just declared?
- Sunk Cost: Has the prospect actually invested yet?
- In-Group: Is the shared identity named + the out-group defined?
- Halo Effect: Does this audience actually respect this authority?
- Confirmation: Does it make the reader feel SMART or DUMB?
- Availability: Are there 3+ concrete sensory details?
- Cognitive Dissonance: Is the system blamed, not the person?