Archives: AI Prompts

  • AI prompts for WooCommerce store owners

    Most WooCommerce catalogues have the same problem: descriptions imported from a supplier, category text nobody read since 2019, and no FAQ where the same three questions repeatedly cost you an order. LLMs are useful for exactly this class of work — volume, structured output, boring but high-margin. Four prompts I actually use for stores.

    1. Rewrite supplier product descriptions without losing the SKU

    You are a WooCommerce copywriter. I will give you a product with a supplier-supplied description, its attributes and its category.
    
    Product name: [NAME]
    SKU: [SKU]
    Category: [CATEGORY]
    Attributes: [color, size, weight, material, etc.]
    Current description (paste as-is): [PASTE]
    
    Rewrite it as:
    1. A one-line "for who and why" (25 words max) that a shopper reads first.
    2. A short paragraph (3–4 sentences) that describes it in the language a real buyer uses, not the manufacturer's.
    3. A bullet list of exactly 5 concrete features — no marketing fluff, only facts a buyer can verify.
    4. A one-line "not for you if" that filters out the wrong shopper (fewer returns is worth more than a marginal sale).
    
    Keep the SKU and the exact attribute values intact. Do NOT invent specifications. If a specification is missing from what I gave you, leave a bracketed placeholder like [WEIGHT?] so I can fill it in.

    Why it works: the SKU + attributes lock the LLM to the actual product, and the “not for you if” line is doing double duty — it reduces returns AND it makes the description sound like a human wrote it, because manufacturers never say who not to sell to. This is the single highest-leverage change on most catalogues.

    2. Generate a FAQ schema block from real support questions

    You are a WooCommerce SEO copywriter. I run a store selling [CATEGORY / PRODUCT TYPE].
    
    Below are 15 real support tickets I received in the past 3 months. Extract the 5 questions that appear most often (paraphrase, do not quote), then write a FAQPage schema section that:
    - Answers each in 2–3 sentences, warmly, from the brand's perspective.
    - Uses the actual language shoppers used, not our internal jargon.
    - Is honest about limitations (return windows, shipping delays, etc.).
    - Ends each answer with a link to the relevant page on the store: [BASE URL].
    
    Output as two blocks:
    1. Clean markdown for me to paste on a product or category page.
    2. A JSON-LD FAQPage block ready to drop into the header.
    
    [PASTE 15 TICKETS HERE]

    Why it works: FAQ schema helps Google, but the real win is on the page itself — the same three questions block half your buyers, and a store owner rarely knows what they are because they are answering them in tickets one at a time. Aggregating actual tickets is the trick that makes this useful.

    3. Upsell copy that does not feel like a used car pitch

    Write me 4 versions of an upsell block that appears on a WooCommerce single-product page.
    
    Main product: [NAME + ONE-LINE DESCRIPTION]
    Suggested upsell: [NAME + ONE-LINE DESCRIPTION]
    Real reason they go together: [ONE HONEST SENTENCE — e.g. "the case protects the screen from the exact drop scenarios that void the warranty"]
    
    Constraints on all 4 versions:
    - Under 40 words each.
    - No exclamation marks. No "Don't miss out!" No "Only 3 left in stock!" if the stock is not actually 3.
    - Each version leads with a different frame: 1) practical, 2) financial (bundle math), 3) social (what other buyers did), 4) preventive (avoids a bad outcome).
    - Include a one-line CTA at the end. Do not use "Add to cart" — write something specific to this product.
    
    Then tell me which of the 4 is most likely to convert for THIS product, and why.

    Why it works: default WooCommerce upsells are noise (“You may also like…”). Framing the upsell around the actual reason the two products belong together makes it feel like advice, not a pitch. The “no false urgency” constraint keeps you out of dark-pattern territory that also happens to hurt long-term trust.

    4. Category text that ranks and reads

    You are a WooCommerce SEO writer. Write a category description for this page:
    
    Category name: [NAME]
    Category slug: /category/[SLUG]
    Number of products: [X]
    Main use-cases (from browsing my catalogue): [LIST 2-3]
    Main audiences: [LIST 2-3]
    
    Structure:
    1. First 40 words: what this category is FOR, in shopper language. This is the meta description.
    2. Next 60 words: how to choose from within the category — what factors matter (price, size, use case).
    3. Final block: a 3-question FAQ that pre-empts the friction points that come up in support (paraphrase, do not fabricate).
    
    Rules:
    - Do not stuff the exact category name into every sentence.
    - Do not use "Welcome to our collection of...". Ever.
    - Include one honest signal about the store's expertise ("we hand-pick", "we test in-house", whatever is TRUE — I will confirm).
    - Include internal links to at most 3 related categories (I will provide slugs on request).

    Why it works: most category text is either empty or scraped. A structure that leads with utility (“what this category is for”) and follows with decision help (“how to choose”) reads like a knowledgeable shop assistant. That translates to lower bounce, more time on site, and gradually better rankings for the category page itself.

    A note on scale

    If you have 500+ products, you are not going to run these prompts one at a time. Wrap them in a script that iterates your CSV export, or use a workflow tool like n8n or Make with the OpenAI/Anthropic node. The prompt is the recipe — the pipeline is up to you. But for a store with 50 products, one focused afternoon with these four prompts and a coffee is enough to lift conversion measurably.

  • Five prompts for sharper positioning (that survive contact with a customer)

    “Positioning” gets thrown around like a magical incantation. What it usually means in practice: someone new visits your landing page, spends 8 seconds, and cannot explain to a friend what you do. Fixing that is a solvable problem, not a Zen quest. Here are five prompts I use before I redesign anything, and after every meaningful shift in the product.

    1. Jobs-to-be-done interview simulation

    You are simulating a Jobs-to-be-Done interview. You are the interviewer.
    
    Product I want to position: [PASTE 3-4 SENTENCES ABOUT PRODUCT]
    Type of customer: [DESCRIBE — role, industry, size]
    
    Simulate a 15-minute JTBD interview with a customer who recently bought this product. Ask, in order:
    1. Take me back to the day you first started looking for a solution. What happened that made you start?
    2. What did you try first, before anything else?
    3. Why did that not work?
    4. When did you first hear about us?
    5. What almost stopped you from picking us?
    6. What is the specific outcome you were hoping for when you signed up?
    7. Two weeks later, what have you actually used it for that you did not expect?
    
    Play both roles. Make the customer's answers specific and self-contradictory in the way real answers are. At the end, summarise the single job the product got hired to do, in the customer's own words.

    Why it works: real JTBD interviews are the gold standard because they surface the “trigger event” that comes weeks before the search, and the actual outcome the customer wanted (rarely the one the product page claims). The LLM simulation is a shadow of the real thing, but 80% of the value at 5% of the cost.

    2. Ten hero-line variations, ranked

    Write me 10 different hero lines for my landing page. My product is [DESCRIBE]. My audience is [DESCRIBE].
    
    Each variation should come from a distinct angle:
    1. What we do (functional)
    2. The transformation the customer experiences
    3. Contrast with the alternative they are currently using
    4. A specific outcome we deliver (with a number if possible)
    5. A pain point framed as a question
    6. A category we are creating or entering
    7. A tribal / identity claim ("Built by ___ for ___")
    8. A punchy opinion ("The old way is dead")
    9. An implicit promise (starts with a verb, ends with a concrete noun)
    10. A pure benefit statement, no jargon
    
    After the 10, rank them best to worst FOR THIS AUDIENCE. Then explain in one sentence per line what would need to be true for the ranking to be wrong.

    Why it works: hero lines are hard because the framing choice matters more than the words. Forcing 10 distinct angles makes you notice which framing your product actually deserves. The ranking + counterfactual step is the useful part — it makes you defend the choice against your own doubts.

    3. Extract the one job from the whole site

    I am going to give you 5 URLs from my site. Fetch them and read them.
    
    [URL 1] — landing page
    [URL 2] — pricing page
    [URL 3] — a comparison or "vs" page
    [URL 4] — a blog post I am proud of
    [URL 5] — the About or Story page
    
    Based on ALL five, and ignoring what any single page says in isolation:
    1. Write one sentence describing the single job this product is hired to do — in the language a customer would use, not the language I use.
    2. List three sentences from the pages above that CONTRADICT that job or muddy it.
    3. Suggest one edit per contradicting sentence that would align it with the primary job.

    Why it works: sites drift. Every page gets written by a different person on a different day, and after 18 months your site says five different things. Reading five representative pages together and extracting the “actual” positioning is a job an LLM is genuinely well-suited to — it has no attachment to any of the pages.

    4. Autopsy a competitor’s pricing page

    You are a pricing consultant. Read my competitor's pricing page at [URL].
    
    Tell me:
    1. What job the pricing page is doing (what decision it is guiding the visitor towards).
    2. Three specific weaknesses — layout, copy, structure, missing information, unclear benefits.
    3. For each weakness, whether it would matter more to a small-catalog buyer or a large-catalog buyer.
    4. The single change to MY pricing page ([MY URL]) that would exploit their weakness.
    
    Be concrete. Name the elements. Do not talk about "visual hierarchy" without pointing at a specific block.

    Why it works: pricing pages age badly. A competitor’s four-plan grid or their over-cluttered feature list is often the fastest positioning gap you can exploit. The LLM is a competent copy editor for someone else’s page — it sees things you cannot because you are too close.

    5. The “explain it to a smart 12-year-old” test

    Take my current landing page (URL: [URL]).
    
    Explain in 3 sentences what my product does, as if you were explaining it to a smart 12-year-old who has never used software.
    
    Then:
    1. Tell me the concept in my landing page that failed the test (something you had to skip or gloss over because it was too abstract).
    2. Suggest a concrete metaphor, analogy, or example I could add to the landing page so the 12-year-old test passes.
    3. Show me exactly what section of the current page you would replace, and what would go in its place.

    Why it works: positioning fails at “abstract nouns you invented” — “AI-powered search intelligence layer” or “programmable observability”. The 12-year-old test forces the language back into concrete verbs and objects. Whatever abstraction the LLM had to skip is the thing your real visitors are also skipping past, just more politely.

    A rule I stole from Bob Moesta

    You are not competing with the other tools in your category. You are competing with the alternative your customer is currently using, which is usually “a spreadsheet, a person, and no formal process”. If your positioning does not clearly beat that specific comparison, no amount of hero-line rewriting saves you. Every prompt above works better when you keep that comparison in mind.

  • Four prompts that turn ChatGPT into an advisor, not a content mill

    Most “AI prompts” articles teach you to produce more — more blog posts, more product descriptions, more cover letters. Fine. The prompts that have actually changed decisions I was about to make do the opposite job: they push back, argue against me, or simulate the outcome I would rather not think about. Four I keep in a text file.

    1. Pre-mortem: assume the plan fails

    You are a senior operator who has seen a lot of plans fail. I am about to commit to this plan:
    
    [PASTE PLAN — 2 to 5 sentences]
    
    Assume that in 6 months this has failed. Write:
    1. The 5 most likely failure modes, ordered by probability.
    2. For each, an early signal I could have detected in the first 4 weeks that would have warned me.
    3. The single change to the plan today that would remove the two biggest risks — even if the plan gets less ambitious as a result.
    
    Be specific. Do not tell me "market conditions changed" — tell me "we underestimated integration time with X and shipped 6 weeks late, and by then the LLM landscape had normalised on Y".

    Why it works: Gary Klein’s pre-mortem is powerful because it unlocks knowledge that “what could go wrong?” does not. Assuming failure has already happened is a different question — the mind goes looking for causes instead of possibilities. LLMs are surprisingly good at this because their training is thick with case studies of shipwrecks.

    2. Adversarial review by a skeptical investor

    You are a partner at a European seed-stage VC firm. You have seen 3,000 pitches, funded 40, watched 30 fail. You are famously skeptical, and you find genuine flaws in almost every pitch.
    
    I just handed you this:
    [PASTE ONE-PAGER, ROADMAP, LANDING PAGE COPY, WHATEVER]
    
    Write the 90-second internal note you would send to your partners after our meeting. Include:
    - The one thing about the founder that impressed you
    - The two things that made you uneasy
    - The single question you would ask that they could not answer well
    - Whether you would take a second meeting, and one honest line about why
    
    Do not be polite. Do not use the phrase "great potential".

    Why it works: sycophancy is the default failure mode of LLMs. Explicitly assigning an unpolite persona — and forbidding polite hedges (“great potential”) — produces feedback that is closer to what a real investor thinks and does not say to your face.

    3. Simulate the customer who almost bought and did not

    Take on the persona of a specific kind of buyer:
    [DESCRIBE — e.g. "a founder of a WooCommerce store with 800 SKUs, doing €40k/month, technically competent but not a developer, in the middle of evaluating search plugins"]
    
    You visited my pricing page ([URL]) and my landing page ([URL]) yesterday. You added the plugin to your compare list. Today you decided NOT to buy. In under 200 words, write me the internal note you would send to yourself explaining why.
    
    Constraints:
    - Be specific about what on the pages triggered doubt.
    - Include one thing you liked so I know what to keep.
    - End with the exact objection your friend would need to answer for you to reconsider.

    Why it works: you get closest to the real objection by simulating the moment right after the decision to walk away, not by asking “what stopped you from buying?” (which invites rationalisation). This prompt catches the actual friction — usually pricing structure, unclear trial terms, or a missing case study.

    4. Identify the load-bearing assumption

    Read this project brief / roadmap / go-to-market plan:
    
    [PASTE]
    
    Identify the single load-bearing assumption — the one belief that, if wrong, invalidates most of the plan. Then:
    1. State it in one plain sentence.
    2. Rate my confidence in it on a scale where 1 is "I made this up over coffee" and 10 is "I have hard data".
    3. Suggest the cheapest experiment I could run in the next 7 days to test whether the assumption holds.
    4. If I fail the experiment, tell me what my next best-guess assumption should be.

    Why it works: plans are usually elaborate structures built on one or two assumptions everyone stopped questioning. Finding it and testing it cheaply is the highest-leverage move you can make in the first week of any project. The LLM is a useful outsider — it has no stake, so it says the thing you and your co-founder have both been avoiding.

    A pattern behind all four

    None of these prompts asks the LLM to generate the answer. They ask it to find the flaw, simulate an outcome, or name the invisible thing. That framing is what separates prompts that change decisions from prompts that fill blog drafts. When the answer feels uncomfortable, you are usually in the right place.

  • Prompts to figure out if LLMs are (already) citing your site

    Being cited by an LLM is the new “ranking on page 1”. You do not get told when it happens, you do not get analytics on it, and by the time you find out someone was asking Perplexity about your niche, the model has already picked its winners for that answer. So the game is: know what LLMs already say about your topic, know who they cite, and figure out what you have to publish to slide into that citation set.

    Below are four prompts I run on GPT-5 or Claude every couple of weeks for my own sites and my clients’. They are prompts, not tips — copy them, paste them, run them. All four assume you have ChatGPT or Claude with web browsing enabled, or you use Perplexity.

    1. Audit whether LLMs know your site exists

    You are a research analyst. I need you to audit how my site is positioned inside your training data and the current web index.
    
    Site: [YOUR URL]
    Topic focus: [ONE SENTENCE — e.g. "semantic search plugins for WordPress"]
    
    Do this:
    1. In one paragraph, describe what my site does based only on what you know. If you do not know, say so.
    2. List 5 queries a real user might send to an LLM about [TOPIC FOCUS]. For each, tell me which 3 sites you would cite in the answer today and why.
    3. For each of those 5 queries, tell me honestly whether my site would appear. If not, name the exact piece of content I am missing.
    4. End with a numbered list of the 3 highest-leverage content gaps I should close to enter the citation set.

    Why it works: LLMs are trained to be helpful, but they do not know if they are hallucinating your site or actually citing it. Asking them to name their sources for realistic queries turns their answer into a citation audit — you see who they cite instead of you, and the gap is the content brief.

    2. Reverse-engineer the citations of a specific competitor

    You are an SEO analyst focused on LLM citations. My competitor is [COMPETITOR URL].
    
    1. Search the web and list 10 topics where [COMPETITOR] gets cited by LLMs today (i.e. shows up in Perplexity answers, appears in ChatGPT responses with browsing, etc.).
    2. For each of the 10 topics, quote the exact fragment or claim the LLM appears to be citing (a stat, a definition, a comparison table, a step-by-step).
    3. Tell me the structural pattern that makes these citations happen (e.g. "comparison table with brand vs brand rows", "definition + FAQ schema", "step-by-step with numbered lists").
    4. Give me a content brief I could hand to a writer to produce the same kind of citation-magnet, but on a topic where MY brand is the authority: [YOUR BRAND, YOUR TOPIC].

    Why it works: LLMs cite structures, not vibes. Comparison tables, honest verdicts, numbered how-to steps and factually crisp definitions get quoted intact. Once you see the pattern the competitor accidentally figured out, you can reproduce it on your own topic in a couple of hours.

    3. Generate a “citation-shaped” article outline

    Write me an outline for a blog post that is optimised specifically to be cited by ChatGPT, Claude and Perplexity — not to rank on Google.
    
    Topic: [YOUR SPECIFIC TOPIC]
    Angle: [WHAT MAKES YOUR TAKE UNIQUE — first-party data, honest comparison, contrarian view]
    
    The outline should:
    - Have a one-line verdict that could be quoted verbatim
    - Include at least one comparison table (rows are options, columns are dimensions)
    - Have H2 sections that answer natural-language questions (e.g. "When is X the wrong choice?" not just "Cons")
    - End with a FAQ section using questions that mirror how someone would talk to an LLM
    - Include specific numbers and facts I can back with a source (list what I need to research)
    
    Do not write the article. Just the outline, section by section.

    Why it works: LLMs quote quotable things. A one-line verdict, a table row, a clear FAQ answer — these are the units LLMs pluck out. Writing to be cited is a different craft from writing to rank on Google, and it starts at the outline.

    4. Track drift over time

    Roleplay as a curious buyer asking ChatGPT for help. Send these exact queries and paste the raw responses back to me, verbatim:
    
    1. "what is the best [TYPE OF PRODUCT] for [SPECIFIC USE CASE]?"
    2. "compare [YOUR BRAND] and [MAIN COMPETITOR] for [SPECIFIC USE CASE]"
    3. "is [YOUR BRAND] worth it?"
    4. "who uses [YOUR BRAND]?"
    
    For each response, tell me:
    - which sources you cited (name the domains)
    - whether my brand ([YOUR BRAND]) appeared, and in what tone (positive, neutral, negative, missing)
    - if missing or misrepresented, what content on my site could change that

    Why it works: LLMs update. Something you published last month might be gaining traction; something a competitor updated yesterday might be pushing you out. Running these exact four queries every 30 days gives you a longitudinal view without any analytics tool being involved.

    One meta-prompt to combine all four

    If you want a single quarterly review instead of four separate runs, ask the model to sequence them: audit → competitor reverse-engineer → outline generation for the top gap → drift check on the queries that surfaced in step 1. Give it 10 minutes and it produces a full quarterly LLM-visibility report. That is genuinely the highest-leverage 10 minutes I spend on any of my sites.