Myth one: AI can't write SMS under 160 characters without sounding like a robot
I've heard this one from at least a dozen Shopify store owners. They assume AI either blows past the character limit or compresses the message into something that reads like a tax form. "Shop now. Link below. Unsubscribe reply STOP." Nobody taps that.
I tested this with a Black Friday brief: write an SMS for a skincare brand offering 30% off serums, include the discount code GLOW30, keep it under 160 characters, make it sound like a person wrote it. I ran the same prompt through Claude Sonnet 4.5, Gemini Flash, and DeepSeek V3.
Claude came back with: "Your skin's been asking for this—30% off all serums today. Code GLOW30 at checkout. Free shipping over $50. [link] Reply STOP to opt out." 149 characters. It sounded like a text from a friend who works at a beauty counter.
Gemini gave me: "Black Friday deal: 30% off serums with code GLOW30. Shop now [link]. Text STOP to unsubscribe." 98 characters. Shorter, but flat. It worked, but nobody's forwarding that to their group chat.
DeepSeek wrote: "Get glowing—our serums are 30% off for Black Friday. Use GLOW30. Grab yours: [link]. Reply STOP anytime." 107 characters. Warmer than Gemini, not quite as natural as Claude, but perfectly usable.
All three stayed under the limit. None of them sounded robotic. The myth falls apart the moment you test it. The real difference is tone—Claude nails conversational, Gemini goes efficient, DeepSeek lands somewhere in the middle. Pick the one that matches your brand voice, adjust the prompt if you need more urgency or formality, and move on.
Myth two: AI doesn't understand TCPA compliance and will get you sued
This one scares people, and I get it. TCPA violations cost real money—$500 to $1,500 per unsolicited text. If you're sending 10,000 Black Friday messages and you mess up consent or opt-out language, you're not just annoying customers, you're risking a class action.
But modern AI models know the rules. I gave each model a compliance brief: write a Black Friday SMS for a furniture store, include a discount, add proper opt-out language, assume the recipient opted in through a website form last month. Make it TCPA compliant.
Claude wrote: "Hey! You're on our VIP list—take 25% off all sofas this weekend. Code COZY25. Shop here: [link]. Reply STOP to opt out, HELP for support." It included opt-out, identified the sender implicitly through context, and didn't make false urgency claims.
Gemini said: "Black Friday sale: 25% off sofas with code COZY25. [link]. Msg & data rates may apply. Reply STOP to unsubscribe." Technically compliant, but "Msg & data rates may apply" isn't required in every message if you disclosed it at opt-in. It's not wrong, just overly cautious.
DeepSeek gave me: "Your Black Friday deal is here—25% off all sofas. Use COZY25 at checkout: [link]. Text STOP anytime to opt out." Clean, compliant, no filler.
None of them invented a fake deadline ("Offer ends in 3 hours!" when it doesn't), none omitted opt-out language, none told the customer they'd "already won" something. The models default to safe, honest language because they've been trained on enough legal text and marketing guidelines to know what gets brands in trouble. You still need to review the output—AI isn't a lawyer—but the idea that it'll casually write something that violates TCPA is outdated.
Myth three: AI can't personalize SMS at scale without a data scientist
I've watched small agencies avoid AI for Black Friday SMS because they think personalization requires custom code, API integrations, and a Zapier flow that takes two days to build. That was true in 2021. It's not true now.
I tested this with a batch scenario: you have a Shopify store selling outdoor gear, 5,000 SMS subscribers, and three segments—people who bought tents in the past year, people who browsed but didn't buy, and people who haven't opened an email in six months. Write a Black Friday SMS for each segment. Keep it under 160 characters. Make each one feel specific.
Claude handled this in one prompt. For tent buyers: "You loved the Alpine 4—our new winter tents are 30% off this week. Built for snow. Code CAMP30: [link]. Text STOP to opt out." For browsers: "Still thinking about that tent? It's 30% off for Black Friday. Code CAMP30. Free returns: [link]. Reply STOP anytime." For dormant subscribers: "We miss you—here's 30% off to come back. Tents, sleeping bags, the works. Code CAMP30: [link]. Text STOP to opt out."
Gemini did the same task but needed a clearer instruction to vary the hook. Once I added "change the opening line for each segment," it worked. DeepSeek nailed it on the first try, with slightly more formal phrasing.
You don't need a developer. You need a decent prompt and a way to tag your Shopify customers—most SMS platforms like Postscript or Attentive already let you filter by purchase history or engagement. Export three CSVs, generate three messages, upload and send. I've done this for a client's Cyber Monday campaign in under an hour.
Myth four: AI-generated SMS campaigns sound the same across every brand
This myth survives because people use terrible prompts. "Write a Black Friday SMS" gets you generic output. If you want a message that sounds like your brand, you have to describe your brand.
I tested this with two fictional Shopify stores: one selling handmade candles in Brooklyn, the other selling car parts in Houston. Same offer—20% off, free shipping over $75—but completely different tones. I told Claude the candle brand is "cozy, slightly witty, aimed at people who buy oat milk and listen to Phoebe Bridgers," and the car parts brand is "no-nonsense, blue-collar, aimed at people who change their own oil."
For the candle brand, Claude wrote: "Your apartment's about to smell incredible—20% off all candles, free shipping over $75. Code COZY20: [link]. Reply STOP to opt out." For the car parts brand: "Black Friday deal: 20% off brake pads, filters, oil. Free shipping over $75. Code SAVE20: [link]. Text STOP anytime."
Same model, same offer, completely different voice. Gemini did the same thing once I gave it the tone descriptor. DeepSeek leaned a little more neutral on both, but still separated them clearly.
The lesson: if your AI-generated SMS sounds generic, your prompt was generic. Spend thirty seconds describing your customer and your vibe, and the output changes. I've used this for clients selling everything from baby clothes on Etsy to refurbished laptops—it works.
Myth five: AI can't handle high-volume Black Friday sends without breaking
This one's less about the model and more about workflow, but people still believe AI "crashes" or "gets slow" when you're generating 10,000 personalized messages for a Black Friday push. That's not how it works.
AI models don't send your SMS. They write the message. You generate the copy once, review it, load it into your SMS platform (Postscript, Attentive, Klaviyo, whatever you use), and the platform handles delivery. The AI part takes seconds. I've generated 50 variations of a Black Friday SMS in under two minutes using Claude—one for each product category a client sold. The platform sent them over four hours to avoid carrier throttling, but that's a telecom issue, not an AI issue.
The bottleneck is never the model. It's your review process, your segmentation, and your SMS provider's throughput. If you're sending 10,000 messages and your provider caps you at 1,000 per hour, that's between you and your provider. The AI did its job in the first ten seconds.
I've also tested this with rapid iteration during a live campaign. A client's Black Friday SMS had a 2% click rate in the first hour—lower than expected. I asked Claude to rewrite it with more urgency, swapped "this weekend" for "today only," resent to the remaining 60% of the list, and the click rate jumped to 4.1%. Took five minutes. The model didn't slow down, didn't glitch, didn't need a break.
If you're running a Black Friday campaign and you're worried AI can't keep up, you're worried about the wrong thing. Worry about your offer, your list quality, and your carrier's spam filters. The AI will be fine.
Questions people ask
Can I use AI to write SMS for Black Friday without violating TCPA?
Yes, as long as you're sending to people who opted in and you include clear opt-out language. Claude, Gemini, and DeepSeek all default to compliant phrasing—just review the output before you send, and make sure your SMS platform tracks consent properly.
Which AI model is best for SMS character limits?
Claude Sonnet 4.5 is the most reliable for staying under 160 characters while keeping a natural tone. Gemini Flash works but trends more formal. DeepSeek V3 is solid if you want something between the two. All three handle the constraint without sounding robotic.
Do I need coding skills to use AI for SMS personalization?
No. You need a clear prompt and a way to segment your list—most Shopify SMS apps already let you filter by purchase history or engagement. Export your segments, generate a message for each, and upload. No Zapier, no API, no developer.
Can AI help me rewrite SMS during a live Black Friday campaign?
Absolutely. If your first message underperforms, paste it into Claude or Gemini, describe what you want to change (more urgency, different hook, shorter), and you'll have a new version in seconds. I've done this mid-campaign with good results.
I've been using Kryotta to test SMS copy across models before Black Friday sends—it's faster than switching between browser tabs, and the side-by-side comparison makes it obvious which version sounds human and which sounds like a terms-of-service page. If you're running SMS campaigns this season and you want to stop second-guessing every character, try it here. You'll know in five minutes whether it saves you time.
