{"slug": "the-best-ai-prompts-for-email-marketing", "title": "The Best AI Prompts for Email Marketing", "summary": "A new library of over 90 AI prompts for email marketing provides copywriters with structured frameworks for generating subject lines, campaign emails, and personalization, organized by campaign type and job-to-be-done. The prompts emphasize specificity over cleverness and include reusable templates for writing custom prompts, with recommendations for which AI models perform best on each task.", "body_md": "# The Best AI Prompts for Email Marketing\n\nMost people prompt AI models for email copy the same way they prompt them for anything else: a single vague sentence, hit enter, copy whatever comes back. Then they wonder why it reads like every other brand's newsletter.\n\nThe prompt is the brief. A weak brief gets you a generic draft. A specific, well-structured brief gets you something close to a finished email on the first try.\n\nThis is a working library, not a listicle. Every prompt below is written to be copied, pasted, and adapted — swap in your brand name, your audience, your offer, and go. We have organised it by campaign type and by job-to-be-done, added a reusable framework for writing your own prompts from scratch, and noted which model tends to handle each task best based on [our own side-by-side testing](/blog/we-asked-10-ai-models-to-write-the-same-email).\n\n## TL;DR\n\n**The best prompts are specific, not clever.** Brand voice, audience behaviour, exact offer terms, word count, and structure beat a well-phrased one-liner every time.**Different models are better at different jobs.** Use ChatGPT or Gemini for structured first drafts, Claude for tone and humanisation passes, Grok for subject line ideation you will then filter.**This library covers 14 campaign types and 90+ prompts**— subject lines, welcome series, win-backs, abandoned cart, promotions, newsletters, post-purchase, re-permission, personalisation, A/B testing, tone rewrites, deliverability checks, and HTML code.**A reusable prompt framework is included** so you can write new prompts for situations this library does not cover.**AI drafts. A human sends.** Every prompt library needs a final human review step — this one has one built in.\n\n## The Prompt Framework: Build Any Email Prompt in 6 Parts\n\nBefore the library, here is the structure behind every prompt in it. Learn this once and you can write a strong prompt for any email situation, not just the ones listed below.\n\n| Part | What to include | Example |\n|---|---|---|\n1. Role |\nTell the model what kind of writer to be | \"You are an email copywriter for a [industry] brand.\" |\n2. Brand voice |\nDescribe or paste examples of your tone | \"Warm, a little playful, never corporate. Here are two emails we have sent that nailed the voice: [paste].\" |\n3. Audience & context |\nWho is receiving this and why now | \"Subscriber has not opened an email or purchased in 90 days.\" |\n4. Offer & constraints |\nExact terms, no ambiguity | \"20% off, code RIDGE20, valid 7 days. No emojis in the subject line.\" |\n5. Structure |\nWhat the output must contain | \"Subject line, preview text, body under 120 words, one CTA button.\" |\n6. Negative instructions |\nWhat to avoid | \"Do not use the phrases 'exclusive offer,' 'act now,' or 'valued customer.'\" |\n\nMiss part 6 and you will get competent, forgettable copy every time — it is the part almost everyone leaves out, and the one that does the most to separate AI-sounding copy from human-sounding copy.\n\n## 1. Subject Line Prompts\n\nSubject lines are worth their own prompts because they carry more weight than any other single line in the email — it is the only part guaranteed to be seen by every recipient.\n\nVolume ideation:\"Generate 15 subject lines for a [campaign type] email to [audience]. Offer: [details]. Vary the approach — include at least 3 curiosity-driven, 3 benefit-driven, 3 urgency-driven, and 3 conversational/casual options. No emojis, no title case, under 50 characters each.\"\n\nCuriosity gap:\"Write 8 subject lines that create curiosity without being clickbait, for an email announcing [product/offer]. Each should raise a question the body answers. Avoid words that commonly trigger spam filters.\"\n\nPersonalisation-style:\"Write 5 subject lines that sound personally written to one subscriber, not broadcast to a list, for a [campaign type] email. Avoid generic openers like 'Hey there' or 'Just checking in.'\"\n\nA/B pair generator:\"Write two subject lines for the same email that test two different psychological angles: one built on loss aversion (what they will miss), one built on gain framing (what they will get). Keep both under 45 characters.\"\n\nPreview text pairing:\"For this subject line: '[subject line]', write 3 preview text options that extend the idea rather than repeating it, each under 90 characters.\"\n\n**Best model for this task:** Grok for volume and creative range, followed by a brand-safety pass in Claude or ChatGPT to filter anything too risky to send.\n\n## 2. Welcome Series Prompts\n\nThe welcome series sets the tone for the entire relationship, so prompts here should establish voice consistency across multiple emails, not just one.\n\nFull sequence outline:\"Plan a 4-email welcome series for new subscribers to [brand], a [industry] company. Email 1 should deliver on the signup incentive. Email 2 should tell our origin story in under 150 words. Email 3 should highlight our 3 bestselling products/services. Email 4 should ask a light engagement question. For each email, give me a subject line, preview text, and a one-sentence summary of the angle — do not write full body copy yet.\"\n\nSingle welcome email:\"Write a welcome email for a new subscriber to [brand]. Tone: [describe]. Include a genuine thank-you, one sentence about what makes us different, and a single CTA to [specific action]. Under 130 words. Do not use the phrase 'welcome to the family.'\"\n\nFounder-voice welcome:\"Write a welcome email as if it is coming directly from the founder of [brand], first person, no corporate 'we.' Mention why the brand exists in one sentence. Keep it under 100 words and end with a question the subscriber could actually reply to.\"\n\nIncentive delivery:\"Write a short welcome email whose only job is to deliver a [discount/freebie] with code [CODE]. Do not oversell — assume the subscriber already wants this. Under 60 words.\"\n\n**Best model for this task:** ChatGPT for the sequence outline and structural consistency across emails, Claude for the founder-voice or narrative-heavy variants.\n\n## 3. Win-Back / Re-Engagement Prompts\n\nThese are some of the hardest emails to write — they need to acknowledge a lapse without sounding needy and create urgency without sounding desperate.\n\nStandard win-back:\"Write a win-back email for [subscriber segment] who has not opened an email or purchased in [X] days. Offer: [details]. Tone: warm, a little playful, not corporate. Under 120 words. Include subject line, preview text, and one clear CTA. No emojis in the subject line.\"\n\nNo-discount win-back:\"Write a win-back email that does NOT offer a discount. Instead, remind the subscriber why they signed up in the first place and ask a single yes/no question about what they are interested in now. Under 100 words.\"\n\nLast-chance / sunset warning:\"Write a final email in a win-back sequence, informing the subscriber this is the last email before we stop emailing them, without guilt-tripping them. Give them one clear, low-effort way to stay subscribed. Under 90 words.\"\n\nSegmented win-back by reason:\"Write 3 versions of a win-back email for [brand]: one for subscribers who stopped after a bad experience (acknowledge without over-apologising), one for subscribers who simply went quiet (light, no assumptions), and one for subscribers who unsubscribed from purchases but still open emails (product-focused, not discount-focused).\"\n\n**Best model for this task:** Claude for tone — win-backs are the campaign type most likely to sound needy or corporate if the model defaults to generic marketing language.\n\n## 4. Abandoned Cart & Browse Abandonment Prompts\n\nTiming and tone are everything here — the first message should help, not sell, and the follow-up earns the right to offer a discount.\n\nFirst-touch cart reminder:\"Write a cart abandonment email for [brand], sent 1 hour after a customer left [product] in their cart. No discount, just a helpful reminder. Include a one-line objection-handler for the most likely hesitation (e.g. shipping cost, sizing). Under 90 words.\"\n\nSecond-touch with incentive:\"Write a second cart abandonment email, sent 24 hours after the first, now including a [X]% discount with code [CODE]. Acknowledge this is a nudge, not the first time we have reached out, without saying so directly. Under 100 words.\"\n\nBrowse abandonment (no cart add):\"Write an email for someone who viewed [product/category] but did not add anything to cart. Do not assume purchase intent — frame it as 'thought you might like these' rather than 'you forgot something.' Include 3 product recommendation slots. Under 90 words.\"\n\nSocial proof cart reminder:\"Write a cart abandonment email that includes one genuine-sounding line of social proof (e.g. review count, bestseller status) about the specific item left in cart, without inventing specific numbers or quotes. Under 90 words.\"\n\n**Best model for this task:** Gemini when connected to real product/inventory data; ChatGPT for a standalone draft without live data.\n\n## 5. Promotional & Sales Prompts\n\nPromotional emails need urgency without desperation and clarity without clutter — the prompt should force a tight structure.\n\nStandard promo:\"Write a promotional email announcing [sale/offer] for [brand]. Audience: [segment]. Include urgency without using the words 'hurry,' 'last chance,' or multiple exclamation points. Under 130 words, one CTA.\"\n\nProduct launch:\"Write a product launch email for [product name], a new [category] from [brand]. Lead with the problem it solves, not the feature list. One clear CTA to [action]. Under 140 words.\"\n\nFlash sale (high urgency, no desperation):\"Write a flash sale email for a [X]-hour sale on [product/category]. Create real urgency using specific time framing (e.g. 'ends tonight at midnight') rather than generic urgency language. Under 100 words.\"\n\nBundle / cross-sell:\"Write an email recommending [product B] to customers who recently bought [product A]. Frame it as a genuinely useful pairing, not an upsell. Under 90 words.\"\n\nPrice increase notice:\"Write an email informing subscribers that prices are increasing on [date], framed around the value they are getting rather than apologising repeatedly. Give them a clear window to buy at the current price if relevant. Under 120 words, transparent and direct tone.\"\n\n**Best model for this task:** ChatGPT for structural completeness on a deadline; Mistral if the sale targets a non-English or European audience.\n\n## 6. Newsletter & Content Email Prompts\n\nNewsletter prompts should prioritise voice and scannability — the goal is to get subscribers reading, not to summarise everything.\n\nRoundup newsletter:\"Write a newsletter intro paragraph for [brand]'s [weekly/monthly] newsletter, teasing 3 pieces of content: [topic 1], [topic 2], [topic 3]. Conversational tone, under 80 words, no corporate 'in this issue' framing.\"\n\nSingle-topic deep dive intro:\"Write an intro paragraph for a newsletter about [topic], written for an audience that already knows the basics. Do not over-explain the topic in the intro — hook them into wanting to read further. Under 60 words.\"\n\nCurated links section:\"Write one-sentence descriptions for each of these 5 links I am including in our newsletter, in our brand voice, that make each one sound worth clicking without being clickbait: [paste links/titles].\"\n\nCommunity/UGC spotlight:\"Write a short newsletter section spotlighting a customer story or piece of user-generated content. Keep the customer's voice authentic — do not oversell it as a testimonial. Under 70 words.\"\n\n**Best model for this task:** Claude for voice-driven content sections; ChatGPT for structured roundups with multiple sections.\n\n## 7. Post-Purchase & Lifecycle Prompts\n\nPost-purchase emails earn future revenue — prompts here should prioritise helpfulness over promotion.\n\nOrder confirmation with personality:\"Write a short order confirmation email for [brand] that has a bit more personality than the default transactional template, without adding unnecessary length. Under 60 words, still includes order number and estimated delivery placeholder.\"\n\nPost-purchase education:\"Write an email sent 3 days after purchase of [product], teaching the customer how to get the most out of it. Not a sales email — genuinely helpful. Under 130 words.\"\n\nReview request:\"Write a review request email sent 14 days after delivery of [product]. Make the ask feel low-effort and specific (e.g. 'one line about X') rather than generic ('leave us a review'). Under 80 words.\"\n\nReplenishment reminder:\"Write a reminder email for a consumable product [product name] that typically runs out after [X] days/weeks, timed to arrive before the customer runs out. Frame it as helpful timing, not a sales push. Under 90 words.\"\n\nVIP / loyalty tier email:\"Write an email informing a customer they have reached [loyalty tier] status. Make it feel earned and specific to their behaviour, not generic congratulations. Include their tier's specific benefits. Under 110 words.\"\n\n**Best model for this task:** Gemini for anything that references specific order/product data; Claude for the tone-sensitive education and review-request emails.\n\n## 8. Re-Permission & List Health Prompts\n\nRe-permission emails must be direct, honest, and never manipulative — the prompt should explicitly forbid guilt-tripping.\n\nRe-permission email:\"Write a re-permission email asking long-dormant subscribers to confirm they still want to hear from us, before we remove them from the list. Make staying subscribed the easy, obvious choice without guilt-tripping them for leaving. Under 100 words, one clear CTA button plus a low-friction 'no thanks' option.\"\n\nPreference center invite:\"Write an email inviting subscribers to update their email preferences (frequency, topics) rather than unsubscribe entirely. Frame it as giving them control, not as damage control. Under 90 words.\"\n\nSunset final notice:\"Write the final email in a sunset policy sequence, informing a subscriber this is their last email unless they take action. Be direct and honest about what happens next, no manipulation tactics. Under 80 words.\"\n\n**Best model for this task:** Claude — these emails fail fastest when they sound manipulative, and tone control is the deciding factor.\n\n## 9. Segmentation & Personalisation Prompts\n\nSegmentation prompts need to keep the core message intact while adjusting tone and proof points for different audiences.\n\nSegment-specific variants:\"Take this base email: [paste email]. Write 3 variants adjusted for these segments: [segment 1, e.g. new customers], [segment 2, e.g. high-value repeat customers], [segment 3, e.g. price-sensitive browsers]. Keep the core structure and CTA the same — adjust tone, proof points, and emphasis only.\"\n\nBehaviour-triggered personalisation:\"Write an email template with placeholder variables for [data points, e.g. recent product viewed, last purchase category, loyalty points balance] that reads naturally with any combination of values filled in. Show me one filled-in example using: [sample data].\"\n\nGeo/seasonal personalisation:\"Adjust this email [paste] for a subscriber in [region/season] — change any seasonal or climate references to be accurate for their context, keep everything else the same.\"\n\n**Best model for this task:** Gemini, consistently the strongest at weaving real data into copy without breaking the narrative flow when connected to an API.\n\n## 10. A/B Testing & Variant Prompts\n\nThe key to useful A/B prompts is isolating one variable at a time — the prompt must enforce that discipline.\n\nAngle test generator:\"Write 2 versions of this email [paste or describe] that test two completely different angles, not just different wording: one built on [angle A, e.g. scarcity], one built on [angle B, e.g. social proof]. Keep length and CTA consistent so the angle is the only real variable.\"\n\nLength test:\"Write a short version (under 50 words) and a long version (under 200 words) of this email [paste], covering the same core offer, so I can test length against my audience.\"\n\nCTA copy test:\"Generate 8 CTA button text variants for an email whose goal is [goal, e.g. get someone to book a demo]. Keep each under 4 words. Avoid generic options like 'Learn More' or 'Click Here' unless specifically requested.\"\n\nTone test:\"Write 2 versions of this email [paste]: one formal and professional, one casual and conversational. Keep the offer and structure identical so tone is the only variable being tested.\"\n\n**Best model for this task:** Any top-tier model works here — the value is in the prompt structure (isolating one variable) rather than the model choice.\n\n## 11. Tone, Rewrite & Humanisation Prompts\n\nThese are the most underused prompts in most workflows — running a second pass on an already-decent draft is often what separates AI-sounding copy from something you would actually send.\n\nDe-corporate rewrite:\"Rewrite this email [paste] to remove corporate marketing language. Specifically replace generic phrases like 'exclusive offer,' 'act now,' 'valued customer,' and 'don't miss out' with more natural, specific alternatives. Keep the same length and offer details.\"\n\nHumanisation pass:\"Rewrite this email [paste] as if a specific person on our team wrote it personally to one subscriber, not a broadcast to a list. Remove anything that sounds templated or generic.\"\n\nCliche audit:\"Read this email [paste] and list every phrase that sounds like generic marketing copy or AI-generated filler. Then rewrite the email replacing each one with something more specific and original.\"\n\nVoice matching:\"Here are 3 examples of emails we have sent that nail our brand voice: [paste examples]. Rewrite this draft [paste new draft] to match that same voice — sentence length, word choice, and level of formality.\"\n\nBrevity pass:\"Cut this email [paste] down to under [X] words without losing the core offer or the CTA. Prioritise cutting throat-clearing sentences and redundant phrases over cutting specific details.\"\n\n**Best model for this task:** Claude, specifically for de-corporate and humanisation passes — it is the most consistent model at avoiding stock marketing phrasing.\n\n## 12. Deliverability & Spam-Risk Prompts\n\nAI models cannot check your sender reputation or authentication records, but they can flag copy-level spam risk before you send.\n\nSpam trigger audit:\"Review this email [paste] for language commonly associated with spam filter triggers — excessive urgency words, all-caps, excessive punctuation, and phrases like 'free,' 'guaranteed,' or 'act now' used in a spammy way. List each risk and suggest a lower-risk alternative.\"\n\nExclamation audit:\"Count the exclamation points in this email [paste] and rewrite it using no more than one, replacing forced enthusiasm with more specific, confident language.\"\n\nLink-to-text ratio check:\"Review this HTML email [paste] and flag if the ratio of image/link content to actual text content looks likely to trigger spam filters. Suggest where to add more text content if needed.\"\n\n**Best model for this task:** Any model can flag obvious issues, but treat this as a supplement to a real deliverability tool, not a replacement — no AI model can see your actual sender reputation or inbox placement data.\n\n## 13. HTML & Email Code Prompts\n\nHTML prompts need explicit constraints for email-client quirks (Outlook, dark mode, accessibility) — the prompt should specify these upfront.\n\nResponsive HTML build:\"Write a responsive HTML email template using inline CSS and table-based layout (for Outlook compatibility) with the following sections: header/logo, hero image, headline, body text, CTA button, footer with unsubscribe link. Optimise for both desktop and mobile clients.\"\n\nConvert copy to HTML:\"Convert this email copy [paste] into a responsive HTML email template with inline CSS, a single CTA button styled in [brand colour], and mobile-safe font sizes. Include alt text placeholders for all images.\"\n\nDark mode safety check:\"Review this HTML email [paste] for dark mode rendering issues — specifically transparent PNGs, hardcoded white backgrounds, and text colours that could become invisible in dark mode clients. Suggest fixes.\"\n\nAccessibility pass:\"Review this HTML email [paste] for accessibility issues: missing alt text, insufficient colour contrast, and CTA buttons that are not screen-reader friendly. List issues and suggest fixes.\"\n\n**Best model for this task:** ChatGPT and Gemini for the most complete, ready-to-use HTML with inline CSS and responsive tables on a first attempt.\n\n## 14. Reporting & Analysis Prompts\n\nReporting prompts are about reasoning, not writing — the prompt should ask for interpretation, not description.\n\nPlain-English performance summary:\"Here are the results of my last 3 email campaigns: [paste open rate, click rate, conversion data]. Summarise the trend in plain English for a non-technical stakeholder, and flag anything that looks unusual.\"\n\nSubject line pattern analysis:\"Here are my last 10 subject lines and their open rates: [paste]. Identify any patterns in what performed well or poorly — length, tone, personalisation, question vs. statement.\"\n\nCampaign post-mortem:\"This campaign [describe] underperformed our benchmark of [X]%. Here is the email copy [paste] and the send data [paste]. Suggest 3 specific hypotheses for what might have caused the underperformance, ranked by likelihood.\"\n\n**Best model for this task:** ChatGPT or Claude — this is a reasoning task more than a writing task, so either top-tier model performs well.\n\n## Common Prompting Mistakes to Avoid\n\n| Mistake | Why it hurts | Fix |\n|---|---|---|\n| No word count | Models default to long, over-explained copy | Always specify a max word count |\n| No negative instructions | You get generic marketing phrases by default | List specific words/phrases to avoid |\n| No brand voice examples | Output sounds like every other brand using AI | Paste 2-3 real emails that nailed your voice |\n| Asking for the \"whole campaign\" in one prompt | Quality drops as scope increases | Break sequences into an outline prompt, then individual email prompts |\n| Never running a second pass | First drafts are rarely send-ready | Add a rewrite/humanisation prompt as a standard step |\n| Treating one model's output as final | Every model has blind spots — hallucinated details, broken merge tags, stiff phrasing | Always apply a human edit before sending |\n\n## Building This Into a Repeatable Workflow\n\nA prompt library is only useful if it turns into a process. Here is a simple version teams can adopt without much overhead:\n\n**Start with the framework, not a blank page.** Fill in the 6 parts (role, voice, audience, offer, structure, negative instructions) before writing a single word of copy.**Draft with a structure-strong model.** Get a complete, correctly formatted first pass.**Run a humanisation pass.** Feed the draft back in with a tone-rewrite prompt to strip generic phrasing.**Check deliverability risk.** Run the spam-trigger and exclamation audits before it goes anywhere near a send list.**Human review.** Check every factual claim, every link, every discount code, and every merge tag by hand. No prompt replaces this step.\n\nFor a deeper look at how different models perform against the exact same brief — including full side-by-side outputs — see our companion piece, [We Asked 10 AI Models to Write the Same Email](/blog/we-asked-10-ai-models-to-write-the-same-email).\n\n## The Bottom Line\n\nThe gap between a mediocre AI email and a genuinely good one is almost never the model. It is the prompt.\n\nSpecific beats clever. A prompt with a real audience, a real offer, a word count, a structure, and a list of phrases to avoid will outperform a beautifully worded one-liner every time — regardless of which model you paste it into.\n\nBookmark this library, adapt the prompts to your brand, and treat every output as a first draft written by a very fast, very well-read intern: talented, occasionally wrong, and always in need of a second pair of eyes before it goes out the door.\n\n## Related Articles\n\n[We Asked 10 AI Models to Write the Same Email](/blog/we-asked-10-ai-models-to-write-the-same-email)[How AI Is Actually Changing Email Marketing Workflows](/blog/how-ai-is-changing-email-marketing-workflows)[The Complete Anatomy of an Email](/blog/email-anatomy-complete-guide)[The Email Metrics That Actually Matter](/blog/email-metrics-that-actually-matter)[Email Marketing Attribution Is Mostly Guesswork](/blog/email-attribution-is-mostly-guesswork)\n\n### Frequently Asked Questions\n\nThe best prompts are specific, not vague. Instead of 'write me a marketing email,' the strongest prompts include the brand name and voice, the audience and their behaviour (e.g. 'has not purchased in 90 days'), the offer and its exact terms, a word count, a required structure (subject line, preview text, body, CTA), and constraints like 'no emojis' or 'no exclamation points.' The more context you give the model, the less editing you will do afterward.\n\nIt depends on the task. ChatGPT tends to produce the most reliably structured, complete draft on a first attempt. Claude tends to produce the most natural, least 'salesy' voice, which makes it strong for tone rewrites and humanisation passes. Gemini is strongest when a prompt needs to incorporate real data such as recent purchases or inventory. Grok is useful for punchy subject line ideation but needs a brand-safety check afterward. Most experienced teams use more than one model in a chain rather than relying on a single model for every task.\n\nAdd explicit negative instructions to your prompt: list the specific phrases you want it to avoid (for example 'do not use exclusive offer, act now, or valued customer'), give it two or three examples of your actual brand voice to imitate, and ask it to write the way a specific type of person would talk, such as 'a friend texting a recommendation' rather than 'a marketing email.' Then run the output through a second, separate 'rewrite this to sound more human and remove any cliches' prompt as a final pass.\n\nYou can reuse a prompt template, but you should not reuse the exact same prompt for every email in a sequence, because the outputs will start to sound repetitive in structure and sentence rhythm. A better approach is to give the model the full sequence outline in one prompt, so it can vary structure and pacing across emails, or to explicitly instruct it to avoid repeating sentence patterns, openers, or CTAs used in previous emails in the series.\n\nYes. This is the single highest-leverage thing you can do to improve AI-written email copy. Paste two or three real emails you have sent that performed well, or a short brand voice guide, before asking for new copy. Models default to generic marketing language when given no reference point, and the difference between a generic AI output and an on-brand one is almost always the presence or absence of real examples in the prompt.\n\n# Time to run those email marketing reports?\n\nLet's get your email marketing reporting set up\n\n[Setup email reporting](/auth/register)", "url": "https://wpnews.pro/news/the-best-ai-prompts-for-email-marketing", "canonical_source": "https://emailcalculator.com/blog/the-best-ai-prompts-for-email-marketing", "published_at": "2026-07-12 21:48:43.090504+00:00", "updated_at": "2026-07-12 21:48:45.520335+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-tools", "natural-language-processing", "large-language-models"], "entities": ["ChatGPT", "Gemini", "Claude", "Grok"], "alternates": {"html": "https://wpnews.pro/news/the-best-ai-prompts-for-email-marketing", "markdown": "https://wpnews.pro/news/the-best-ai-prompts-for-email-marketing.md", "text": "https://wpnews.pro/news/the-best-ai-prompts-for-email-marketing.txt", "jsonld": "https://wpnews.pro/news/the-best-ai-prompts-for-email-marketing.jsonld"}}