Today’s Marketing

AI marketing prompts

The prompts I post on X, in one place. Copy one, paste it into your AI with your own files.

Updated · By Blake Emal

  1. An AI agent that briefs you every morning

    how i'd set up an AI agent that briefs me every morning
    
    (competitors, reviews, forums, all in one message at 8am)
    
    1. pick the sources (30 min)
    
    - competitor websites, changelogs and blogs
    - competitor reviews on the review sites your buyers use
    - forum and X searches for your category and competitor names
    - your own new reviews and support tickets
    
    2. build the agent
    
    use an always-on AI agent, or an automation on a daily schedule:
    - check each source every morning
    - keep anything new since yesterday
    - send it all to the AI with the brief prompt below
    
    3. the brief prompt
    
    "you're my market analyst. from today's new items, write a brief under 300 words:
    - what changed at each competitor (pricing, features, messaging)
    - new complaints about competitors, with quotes
    - new questions buyers are asking
    - one thing we should do this week because of this
    skip anything that isn't new. link every item"
    
    4. deliver it where you'll read it
    
    team chat, email or a text at 8am. one message, same time every day
    
    5. use it
    
    - every competitor complaint → a post on how you handle it
    - every new buyer question → a post or a help article
    - every competitor change → a note in your positioning doc
    
    6. review it monthly
    
    ask the agent to read the last 30 briefs and summarize the 3 biggest shifts in your market
  2. Check which pages AI cites for your category

    run this in your AI once a month, web search on:
    
    "what are the best [category] tools for [buyer]? list your sources"
    
    if you're not in the sources, those pages are your content plan
  3. Turn a post that did 3x your median into 5 more

    post did 3x your median? paste it into your AI
    
    ask for 5 new posts with the same structure on different topics
    
    a structure that worked once usually works again on a new topic
  4. A month of customer research in one afternoon

    how i'd do a month of customer research in one afternoon w/ AI
    
    1. collect (30 min)
    
    - your reviews + your top 3 competitors' reviews
    - support tickets, sales call transcripts, DMs
    - AI with web search: "find forum threads where [buyer] asks for help with [problem] in the last year. list links"
    
    dump everything into one AI chat
    
    2. cluster (20 min)
    
    "group all of this into the 5 biggest problems. for each: how often it appears, 3 exact quotes, and which type of customer has it worst"
    
    3. price the pain (20 min)
    
    "for each problem, find any mention of money, hours or risk. estimate what it costs the customer per month and show the quotes you used"
    
    4. find the gap (20 min)
    
    "what do people ask for that none of these companies offer? what do they say they've tried that failed?"
    
    5. find the words (20 min)
    
    "list 20 phrases buyers use that a marketer would never write. keep their spelling"
    
    6. turn it into decisions (1 hour)
    
    one page with:
    - the problem worth leading with (highest frequency x highest cost)
    - the segment that has it worst
    - 5 phrases for your site and ads
    - 3 things to build or offer next
    
    7. ship something the same week
    
    rewrite your top-of-page headline with the buyer's words and A/B test it against the current one
    Receipt: run on Buttondown's customers page

    Step 5 ("list 20 phrases buyers use that a marketer would never write. keep their spelling"), run on the quotes visible on Buttondown's Customers page as captured. Four quotes only gave 7 phrases.

    1. "NEXT LEVEL AMAZING" (about customer service)
    2. "SO MUCH SIMPLER to use its backend"
    3. "It just keeps getting better and more delightful"
    4. "I couldn't be happier with it"
    5. "sending out irregular updates to a few dozen friends"
    6. "It is absolutely a joy to use"
    7. "the best support I have ever encountered"
    
    Use number 5. It names a real situation instead of a feeling, so a headline built on it finds the exact person who has it.
    See it in the library
  5. 7 prompts to hear what your buyers want

    steal these 7 prompts to hear what your buyers actually want:
    
    (start one AI chat, upload your reviews, call transcripts, support tickets and DMs. then run these in order)
    
    1. "list the 15 phrases these customers use most to describe their problem. quote them word for word. no paraphrasing"
    
    2. "what happened right before each person went looking for a solution? group the triggers and rank them by how often they appear"
    
    3. "what are these buyers afraid will happen if they buy? list the fears they hint at but never say directly. quote the hint"
    
    4. "besides other products, what are these people using today instead? include spreadsheets, freelancers and doing nothing"
    
    5. "which outcomes do these customers tie to money, time or status? rank them by frequency and quote one example each"
    
    6. "here are my competitor's worst reviews. write the one promise that would make each of these people switch to us"
    
    7. "write 10 one-sentence pitches using only words from these customers. score each 1 to 10 on how specific it is and explain the score"
    
    What did I miss?
    Receipt: run on Tuple's love page

    Prompt 7, run on the 6 quotes on Tuple's love page as captured. Six short quotes only had enough words for 4 pitches.

    1. "Best screen sharing tool ever, for when Slack miserably fails at screen sharing." 9/10. Names the moment and the tool it beats.
    2. "Brings new people up to speed more quickly and keeps your velocity high over time." 7/10. A clear result, but no number.
    3. "It Just Works and gets completely out of your way so you can focus on the task at hand." 5/10. Most tools could say this.
    4. "Pair programming is a superpower for remote teams." 4/10. Sounds good, says little.
    
    The best pitch came from a complaint about another tool, not from the praise.
    See it in the library
  6. Test an offer on 3 buyer personas

    before you launch an offer:
    
    paste 30 real customer quotes into your AI, ask it to build 3 buyer personas from those quotes only, then pitch each one the offer
    
    ask "what would stop you from buying?"
    Receipt: run on Baremetrics' Wall of Love

    The 3 customer quotes on Baremetrics' Wall of Love as captured (Contactually, FullContact, Subbly; the Subbly quote is cut off in our capture). Personas from those quotes only. The "what would stop you" answers are the AI's guesses.

    1. The spreadsheet escapee (Contactually). Used to "wait days or weeks to chug through the data in spreadsheets."
    What would stop them: "Will these numbers match the ones I already trust in my sheets?"
    
    2. The builder with no spare time (FullContact). Wants to stay "focused on building our product" with "no code."
    What would stop them: "How long does setup really take, and who keeps it running?"
    
    3. The decision maker (Subbly). Wants "the insight necessary to make better business decisions."
    What would stop them: "Will it tell me what to do, or just show me more charts?"
    
    What to test: answer each fear on the page, right next to the quote from that kind of buyer.
    See it in the library
  7. Find the customer segment that pays most and stays longest

    Paste this into your AI with your customer export:
    
    "You are a growth analyst. Attached is a CSV of every customer from our payment processor: signup date, plan, total paid, status (active or canceled), and company or email domain.
    
    1. Group customers into segments by company type, size or use case. Infer from the domain and plan where needed and say when you're guessing
    2. For each segment: number of customers, average total paid, % still active, average months before canceling
    3. Rank segments by total revenue and by retention
    4. Name the one segment that pays the most AND stays the longest
    5. For that segment, write: who they are in one sentence, the likely reason they stay, and 3 places they probably spend time online
    6. Name the segment that churns fastest and the most likely reason
    7. Recommend one change to who we market to next month, based only on this data
    
    Show your math in a table before your conclusions. Don't invent numbers that aren't in the file."
    Receipt: run on Linear's pricing page

    No company posts its customer export, so this is step 1 only: group buyers by plan from Linear's pricing page as captured Oct 20, 2026, and say when it's a guess.

    Free ($0): 2 teams, 250 issues. Guess: small teams trying it. The 250 issue cap is where they hit the wall.
    Basic ($10 per user/month, billed yearly): 5 teams, unlimited issues, admin roles. Guess: small teams that hit the cap.
    Business ($16 per user/month): unlimited teams, Code Intelligence, Zendesk and Intercom integrations. Guess: teams that tie product work to code and support.
    Enterprise (custom, annual billing only): SAML and SCIM, invoice/PO billing. Fact from the page: built for security and procurement checks.
    
    Steps 2 to 7 can't run on a pricing page. Who pays most and stays longest needs your own export.
    One thing the page does show: Business is the only plan with two buttons and a filled-in "Get started." That's the plan Linear is pushing.
    See it in the library
  8. Ask your sales page "what's the catch?"

    paste your sales page into AI and ask "what's the catch?"
    
    your buyers can run the same check. fix whatever it finds before they do
    Receipt: run on Cursor's pricing page

    The prompt as written, run on Cursor's pricing page as we captured it on Oct 20, 2026 (Individual tab, top of the page).

    1. Free says "Limited" agent and chat requests. There's no number, so you can't tell when you'll hit the wall.
    2. Plus gives you a "Shared AI usage pool." The page doesn't say what it's shared with.
    3. Premium is "4x more usage compared to Plus" and Super is "15x." Plus never gets a number, so 4x of what?
    4. Super adds "Priority model access at peak times." That hints the cheaper plans can be slower when it's busy.
    
    Fix first: put one real number on Plus. Every other plan is priced off it.
    See it in the library

  9. Find what your best posts have in common

    steal this:
    
    paste your last 30 posts into your AI with impressions and saves next to each one
    
    ask "what do the top 5 have in common that the bottom 5 don't? be specific about the first line"
    Receipt: run on 14 brand posts from our viral X collection

    We don't have saves for other brands' posts, so this ran on views. 14 posts from the viral X post patterns collection (X API counts), top 5 vs bottom 5, first lines only.

    Top 5 (1.1M to 4.2M views):
    Fortnite: "Tempting, isn't it?"
    MrBeast: "I need product to fill a giant store for a top secret video."
    Xbox: "How our fall will be looking"
    McDonald's: "3 boxes. 15 toys. 2 cards with every collectible."
    Tesla: "What goes up must come down"
    
    Bottom 5 (91K to 426K views):
    Netflix: "HUNTR/X don't miss."
    Duolingo: "i heard they're selling me today?"
    DoorDash: "BALLHALLA"
    Samsung: "Today's essentials, j-hope edition"
    Wendy's: "we got a wendys video game before gta vi"
    
    The top 5 make you look at the video or image to finish the thought, or they give you facts you can count. The bottom 5 only land if you already know the reference.
    
    Caveat: account size drives views here more than the line does. On your own posts the account stays the same, so the first line is what changes.
    See it in the library

Bring Blake what the prompts found

50 minutes on Google Meet. We work on it together and you leave with the new version written in a doc you can use the same day. $499.

Book my call

One email every Monday

The week's nine best examples, each with how to do it yourself.