Give Claude your website and a competitor's. Find the moments where they make buying easier, then turn those findings into changes you can actually use.
- Find why a customer might choose them over you
- Compare websites through three customer perspectives
- Break down competitor ads, offers, reviews and content
- Turn the findings into original copy and a practical test
- 16 complete prompts + your one-page quick reference
Start Here: Think Like Your Customer
The plan
- 1Set up your competitor research with the master prompt.
- 2Compare websites through your customer's eyes.
- 3Break down ads, offers, reviews and content.
- 4Find a reason for someone to choose you.
- 5Turn the evidence into better copy and one useful test.
Before you start
Your business
Your website, who you help, your offer and one thing you want to improve.
Competitors
One to three businesses a customer would realistically consider instead of yours.
Claude
A chat at claude.ai, plus public links or readable screenshots. No coding required.
You know your business. Your customer is meeting it for the first time. Use Claude to find the questions your competitors answer more clearly, then improve your own explanation.
Set this up in a normal Claude chat
- Open Claude and start a chat. Claude Code and Cowork aren't required.
- Paste the master prompt below. Replace the square brackets and answer its questions. Voice notes are fine.
- Ask it to search the web. The new experience searches when useful. If your version has Web search under the + menu, enable it.
- For blocked pages, use +, then Add files or photos, to upload screenshots. Include the source URL and capture date. Pasted text works too.
Start with one competitor. Use the follow-up prompts in this chat. Bring the master prompt and your saved summary if you start a new chat later.
Maverick's tip
Customer simulations aren't real interviews. Public ads don't show profit. Use the research to choose what to test, not to manufacture certainty.
Paste Your Master Research Prompt
This sets up the whole process. Claude should ask about your business before handing you a list of generic marketing advice. Copy the entire box, including the last instruction.
Prompt 1: Build my competitor research plan
Be my practical competitor-research partner. Help me understand why a customer might choose another business, then turn what we find into specific improvements to mine. My business and website: [NAME + URL]. What I sell and typical price: [OFFER + PRICE, OR UNKNOWN]. My ideal customer and location: [CUSTOMER + LOCATION]. Competitor websites: [ONE TO THREE URLS, OR HELP ME CHOOSE]. My main goal: [MORE QUALIFIED INQUIRIES / MORE PURCHASES / CLEARER POSITIONING / BETTER ADS]. My constraints: [TIME, BUDGET AND THINGS I CANNOT PROMISE]. First ask up to five questions about my customers, buying objections, strengths, current results and goal. Wait for my answers. Then suggest a focused research plan and let me confirm it. Use current public pages and materials I supply. Read the actual pages, not just search snippets. Record the URL or filename, capture date, exact page or screenshot, and relevant evidence. Separate observed facts, your interpretation and things we still need to check. If access fails, ask for screenshots or pasted text. Don't fill gaps with guesses. Compare the customer journey: ad or search result, website, offer, proof, price and next step. Look at what my competitors do better AND where my business has an advantage. Use the same criteria for everyone. A missing detail means not found in the reviewed material, not that the business doesn't offer it. Customer simulations are hypothetical. Do not invent sales, conversion rates, ad spend, profitability, testimonials or customer opinions. Flag small or biased samples. Don't turn my competitor's claims into verified facts or copy their wording. Give me: a short evidence table, five useful findings, three changes worth making, draft copy using confirmed facts, and one way to test the biggest assumption. Explain everything in plain English. Keep source links beside the findings they support. Work in stages. Start with your questions now. Don't publish, contact anyone, submit forms, buy anything or activate an ongoing task.
Give Claude Evidence It Can Actually Use
You don't need to upload an entire website. Gather the pages that answer a customer's buying questions: the homepage, the relevant product or service page, prices, FAQs and what happens after they click the main button.
If you're comparing design, add screenshots. Reading page text doesn't prove Claude saw the layout. For video ads, supply the transcript and a few key frames if it can't inspect the video itself.
Keep a simple research folder
Save each item with a useful name, like competitor-a-pricing-2026-09-30.png. A text note beside it can hold the source link and what you captured. The date matters because offers, prices and ads change.
| Collect this | Include this detail |
|---|---|
| Website page | Full URL, page name, capture date, relevant screenshot or text |
| Ad | Advertiser, ad link or ID, platform, visible dates, creative and landing page |
| Review | Source, date, rating if visible and relevant wording |
| Your own business facts | Real prices, service area, turnaround, guarantees and proof you can substantiate |
Prompt 2: Check my research material
Review the URLs, screenshots and notes I've supplied for my business and [COMPETITOR]. Before analyzing them, make an evidence inventory: item, source URL or filename, date, what you can actually read, and what's missing. Identify mismatched comparisons, such as different products, countries, billing periods or customer types. Separate website claims from independently supported facts. Tell me which three missing items would most improve this comparison, and why. Ask for those instead of pretending you accessed them. Then create a compact evidence table we can reuse. Give each item a short ID so later recommendations can point back to it. Don't infer that an unavailable page or missing detail doesn't exist.
Maverick's tip
A screenshot should include the words Claude needs to read. Upload a close crop as well as the full page if the text is tiny. Check that Claude read the actual price and conditions correctly before continuing.
Find the Gaps on Your Website
Most businesses ask AI to make their homepage sound better. This gives it a more useful job: explain where a potential customer still has questions, and whether a competitor answers them more clearly.
Prompt 3: Compare the buying experience
Using our business brief and evidence table, compare my website with [COMPETITOR]. Focus on the specific product or service our shared customer would buy. Compare: who it's for, the promised outcome, what's included, price clarity, proof, likely objections, the main call to action and what happens next. For each, show my page detail, their page detail, and why the difference could matter to a buyer. Cite the relevant URL or evidence ID. Mark anything you couldn't inspect. Find five specific things they communicate better and three advantages I should preserve. Don't force the counts if the evidence doesn't support them. Separate a better-written claim from a genuinely better offer. Don't give arbitrary numerical scores. Finish with the single most confusing moment on my site, the simplest useful fix, and one question I should ask a real customer to check your interpretation. Don't rewrite the whole website yet.
What a useful finding looks like
Fictional example: You run a cleaning business. Your site says, "Contact us for a quote." Another site lists the rooms included, explains how the quote works and shows the next available booking window.
The useful finding isn't "their website is better." It's "a customer can work out what happens next without having to contact them first."
Your fix might be a short "How booking works" block. You may already have a great process. You just haven't explained it where someone needs to see it.
Maverick's tip
Ask Claude to point to the exact sentence, section or missing answer behind every recommendation. "Build more trust" isn't specific enough to act on.
Let Three Different Customers Compare You
A customer watching every dollar makes a different decision from someone who needs help tomorrow. Use Claude to explore those differences instead of pretending everyone buys for the same reason.
Prompt 4: Simulate the customer's decision
Based on our brief, create three plausible customer profiles with different buying priorities. Use what I've told you about real customers, and mark anything assumed. Include a price-sensitive buyer, an urgent buyer and a cautious buyer only if they fit this business. For each, simulate comparing my site with [COMPETITOR]. Walk through: what catches their attention, what they understand, what builds trust, what creates doubt, what they still need to know, and which business they'd tentatively choose. They can choose neither or remain undecided. Explain the page details behind each reaction using our evidence IDs. Label every decision as a hypothetical simulation, not an actual customer finding. Don't invent demographic precision, quotes from real people or conversion probabilities. Finish with one change that helps each profile and one neutral interview question to ask a real customer. Then challenge the strongest recommendation: what missing evidence could reverse it?
| Hypothetical buyer | The question they need answered |
|---|---|
| Watching their budget | What will this actually cost, and what's included? |
| Needs help quickly | When can you do it, and how do I get started? |
| Has been disappointed before | Why should I trust that you'll deliver? |
Check one assumption with a real person
Show a customer the page without explaining it first. Ask, "What would you need to know before taking the next step?" Let them answer before showing Claude's prediction. You're looking for the thing they notice on their own.
Maverick's tip
A confident simulated buyer is still a simulation. Keep the interesting questions it raises, then check the important ones with people who actually buy this kind of thing.
Find the Ads Your Competitors Are Running
Ads show how a business introduces itself before a customer ever reaches its website. They can reveal which problems, offers and objections that business chooses to emphasize.
Start with Meta Ad Library
- Open Meta Ad Library.
- Choose the relevant country and the All ads category. Search for the business or advertiser name, then check that you've found the right page.
- Save a small selection of distinct ads. Capture the creative, copy, visible dates, ad link or ID, and the destination page. You don't need twenty nearly identical versions.
Then check Google
Open Google Ads Transparency Center. Search by advertiser name or website. Check the advertiser identity, then use the date and location filters that fit your comparison. Save examples and the visible context, just as you did for Meta.
If Claude can't access a library page, collect the material yourself and upload it. Empty results can reflect your filters, the advertiser name or coverage. They don't prove the business isn't advertising.
Prompt 5: Organize my competitor ads
I've supplied public ad examples for [COMPETITOR], with links or screenshots and capture dates. Organize the examples by distinct creative concept, grouping near-duplicates so they don't distort the picture. For each concept, record the platform, evidence ID, opening hook, visible offer, intended audience clues, main visual, call to action and landing page. Separate observations from guesses. Tell me when a video transcript, frame or destination is missing. Summarize the three messaging themes in this sample. State the sample size and what it doesn't cover. Don't label anything a winning ad or infer spend, click-through rate, conversions or profit from its age or number of versions.
Maverick's tip
An ad that's been running for a long time can be worth studying. Its age alone doesn't tell you whether it works. Public libraries are research material, not access to a competitor's results dashboard.
Turn Ad Research Into Original Ideas
You're looking for why an ad makes sense to a customer. Maybe it starts with a frustrating problem. Maybe it shows the result instead of talking about it. Maybe it removes the reason someone keeps putting off buying.
Take the underlying idea and make it specific to your business. Don't ask Claude to swap the brand name in someone else's ad.
Prompt 6: Find the useful idea behind each ad
Analyze the ad sample we organized. For each distinct concept, explain the customer problem, promised outcome, opening hook, proof offered, objection addressed and next action. Cite the actual copy or visual detail. Distinguish what the advertiser claims from what the evidence proves. Then identify three opportunities for my business: an unanswered question, a customer situation not covered, or a strength we can demonstrate honestly. Check each against my confirmed capabilities. Create five ORIGINAL ad concepts for me. For each, give the target situation, opening line, visual or shot idea, short body copy, call to action, matching landing-page promise and evidence I'd need to support the claim. If a fact is missing, ask me rather than invent it. Do not copy their wording, distinctive creative or customer testimonials. Treat these as ideas to test, not predicted winners. Finish by choosing one concept to develop first and explain that choice using our goal and available proof.
A simple way to make this less generic
Fictional example: Every cleaning ad in your sample says "Get your weekends back." You also know your customers worry about who will enter their home.
An original angle could show your real team, explain who arrives and walk through your actual first-visit process. It answers a different buying question. Only use the team checks and service promises your business genuinely provides.
Maverick's tip
If Claude gives you five versions of "Save time and money," ask it to name a specific moment in the customer's day for each one. The situation should change, not just the adjective.
Follow the Ad All the Way to the Page
An ad can be clear and the page can still lose the customer. The offer disappears, the price changes, or the button asks them to do something they weren't expecting.
Pair the ad with its actual destination. If the destination redirects, record the page you really land on. Inspect the visible journey without submitting forms or making purchases.
| The ad | The page | The next step |
|---|---|---|
| The promise they click | The proof they find | What happens now |
Prompt 7: Find the broken promise in the journey
Compare [AD EVIDENCE ID] with its destination [LANDING PAGE URL OR SCREENSHOTS]. Then compare my equivalent ad and page, if supplied. List the expectation the ad creates, what the page confirms, what changes or disappears, what questions remain, and what the next step asks the visitor to do. Check the offer, audience, pricing conditions, proof and call to action. Cite each detail. Don't claim to have tested a form or checkout unless I've supplied that evidence. Give me three likely points of confusion, with the simplest fix for each. Use language like "could cause doubt," not invented conversion losses. Draft one opening section for my page that follows through on my actual ad promise, using only confirmed facts. Finish with a manual checklist I can follow to confirm the experience myself. Stop before submitting anything or changing the live page.
The question to keep asking
"If I clicked because of this promise, would this page immediately make sense?" That's a more useful first check than asking Claude to redesign every element on the page.
Maverick's tip
Compare the same audience and offer. A free consultation ad and a paid product checkout might serve different goals even if the businesses compete.
Compare Offers Without Racing to the Bottom
The lowest price doesn't automatically make the clearest or strongest offer. One business may include setup, another may charge for it, and a third may require a longer commitment.
Ask Claude to compare what someone actually gets and what they have to agree to. Keep unknown prices marked unknown. A "starting at" price is not a final quote.
Prompt 8: Find the difference in the offer
Compare my offer [DETAILS] with [COMPETITOR OFFER LINKS OR FILES]. Build a side-by-side view of the intended customer, result promised, included work, exclusions, price basis, currency, commitment, onboarding, support and stated guarantee or cancellation terms. Use the same product scope and billing period where possible. Explain mismatches instead of forcing an apples-to-apples total. Cite the source and date for prices and terms. Label starting prices, promotions, missing taxes or extra costs only when the evidence supports them. Don't invent a quote. Identify where their offer is easier to understand, where mine is stronger, and what a buyer still needs to ask. Suggest three improvements to how I explain or package my offer without automatically lowering the price. For each proposed change, tell me what my business would need to deliver, what I must confirm, and a plain-English way to describe it. Don't invent guarantees, availability or included services. Finish with a short comparison checklist a customer could actually use.
Make the improvement practical
Fictional example: If your cleaning price already includes supplies, say so. If a customer must provide parking, explain that before booking. Clarity about the small details can be more useful than a new discount.
Maverick's tip
Separate a wording change from an operational change. "Explain our existing arrival window" is copy. "Promise same-day service" changes what your business must deliver.
Read Reviews for Unanswered Customer Needs
Reviews can show you the words people use when they're happy, frustrated or surprised. They can also show what customers expected but didn't receive.
Collect a balanced sample from the same general period. Include positive, mixed and negative reviews instead of only the complaints that support your argument. Record the source, date and rating if visible. Remove unnecessary personal details before uploading.
Prompt 9: Find the patterns in customer feedback
Analyze the review sample I've supplied for my business and [COMPETITOR]. Use only these reviews unless I ask for more research. First report the source, date range, number of reviews per business and important gaps in the sample. Group feedback into concrete themes: reasons for choosing the business, moments customers value, recurring disappointments, buying objections and surprises. Count themes within THIS sample, showing the denominator. A review can mention more than one theme. Identify possible duplicates and uncertainty. Use short attributed excerpts with review IDs where useful. Don't invent quotations or treat selected reviews as representative of all customers. Distinguish allegations and opinions from independently verified facts. Give me three needs we could address honestly, three phrases that help us understand customer language, and five questions to ask our own customers. Turn one confirmed strength into draft website copy without reusing someone else's review as our testimonial. Keep unsupported conclusions clearly labeled.
Look beyond the star rating
"They told me exactly when they were arriving" can be more useful than "Great service." It tells you what made the customer feel looked after. If your business does that too, make it visible in the buying process.
Maverick's tip
If your sample has 18 negative reviews and two positive ones, Claude should flag that imbalance. It shouldn't announce that 90% of the competitor's customers are unhappy.
Find the Content Questions They Miss
You don't need to copy a competitor's posting schedule. Look for the questions customers ask before buying, then see which ones the competitor answers clearly and which ones remain confusing.
Use a defined sample, like their last ten relevant posts and five useful website articles. Include dates, links and transcripts where you have them. Visible likes and views are context, not proof that a post produced sales.
Prompt 10: Build a useful content gap plan
Review the competitor posts, articles and transcripts I've supplied, plus our customer questions and business brief. State the sample size and date range. Group the content by the customer question it answers and the stage it serves: understanding the problem, comparing options, deciding to buy, or getting value afterward. Identify useful explanations, repeated topics and questions absent from this sample. Don't claim a competitor has never covered a topic unless we've actually checked. Give me six original content ideas tied to genuine buying questions for my business. For each, include a working title, specific audience situation, three points to explain, an example or demonstration I can produce, and a sensible next step for the reader. Prioritize three using customer relevance and evidence we already have. Don't promise rankings, reach or revenue, and don't use public engagement numbers as a substitute for conversions. Finish by drafting the opening of the strongest idea in my voice.
Questions that often lead somewhere useful
What does this cost? Who isn't it right for? What happens on the first day? What should I prepare? What goes wrong when someone chooses the cheapest option? What does a good result actually look like?
Maverick's tip
Ask Claude which idea you could demonstrate with something you already have. A real example from your work is usually more convincing than a general list of advice.
Give People a Clear Reason to Choose You
After all that research, you should have more than a list of things to imitate. Look for a useful strength you can explain and prove, especially if competitors don't emphasize it in the material you reviewed.
That might be a specific type of customer, a clearer process, specialist experience or a service detail people care about. "Better quality" needs an explanation.
Prompt 11: Find a defensible position
Using our evidence and my confirmed business strengths, propose three clear reasons a suitable customer might choose us. For each, identify the customer, the situation they face, our specific advantage and the proof we can actually show. Compare the idea with the competitor material we've reviewed. Distinguish a capability they lack from something they simply haven't emphasized in this sample. Don't call us unique, best, fastest or the only option without enough evidence. Write one positioning sentence, one homepage headline and one supporting proof idea for each direction. Keep the language specific and human. Explain who the direction might not appeal to and what assumption needs checking. Recommend one direction based on my goal and available proof. Then make the strongest case AGAINST it, and give me one question to ask a real customer before choosing it.
An example of making the claim concrete
Vague: "Cleaning you can trust."
More specific, if true: "A regular cleaning team, a clear arrival window and a checklist you can review before the first visit."
The second version gives the customer something they can understand and check. Your proof could be a real checklist, a process explanation or an approved customer example.
Maverick's tip
Don't pick a position just because no competitor says it. It still needs to matter to a customer and be something your business can deliver.
Turn the Findings Into Better Copy
Now ask for the words. Claude has your offer, audience, evidence and possible buying objections, so it has a much better starting point than "make my website better."
Give it a short sample of writing that sounds like your business. Choose one page first. Keep changes tied to the customer question you're trying to answer.
Prompt 12: Write the changes I can actually make
Using our research and the positioning direction I've chosen, draft improvements to [MY PAGE URL OR TEXT]. My writing sample is [PASTE SAMPLE]. My confirmed offer facts are [FACTS OR REFERENCE OUR BRIEF]. Focus on these three customer concerns: [CONCERNS, OR ASK ME TO CHOOSE FROM OUR FINDINGS]. For each change, show the current wording or missing section, the proposed wording, where it belongs and the evidence behind the recommendation. Include a homepage headline, supporting paragraph, main call to action, a short "what happens next" section and five FAQs. Keep the tone natural and specific. Make the button accurately describe the next step. Do not copy competitor wording or invent numbers, results, testimonials, prices, scarcity or guarantees. Mark missing facts as [CONFIRM] and list the questions I need to answer. Don't put a placeholder into copy labeled ready to publish. Finish with a short review checklist. Check that the new copy still matches what my business actually delivers. Give me drafts only; don't publish or change the website.
Read the finished copy out loud
If it sounds like every other business in your industry, ask Claude to make the claims more concrete. If it sounds like a promise your team can't keep, fix the promise before polishing the sentence.
Maverick's tip
Keep your old version. Save the proposed copy with the date and the reason for the change so you can tell later whether it helped.
Choose One Change and Test It
You don't need to rebuild your entire business because Claude found twelve possible improvements. Choose a change with a clear reason behind it and a result you can observe.
If you have very little traffic, start with real customer conversations and a simple before-and-after review. A tiny website A/B test may take too long to tell you anything useful.
Prompt 13: Make my seven-day action plan
Turn our findings into a realistic seven-day plan. I have [HOURS] available, [BUDGET] to consider and this current baseline: [KNOWN RESULTS, OR UNKNOWN]. Don't assume I can spend money or publish without reviewing the plan. Choose three actions at most. For each, state the customer problem, supporting evidence, exact change, effort, dependency and a useful result to watch. Separate copy-only fixes from changes that require new service delivery or proof. Order them by likely usefulness and strength of evidence, without pretending the ranking is measured fact. Design one test for the highest-priority action. Specify what changes, what stays the same, the main measure, a quality check and how we'll collect the evidence. If traffic is too low for a meaningful A/B test, suggest a customer interview or usability check instead. Tell me what would count as encouraging, inconclusive or a reason to stop. Don't invent a sample size, guaranteed uplift or statistical significance. End with the first task I can complete today and the draft material I need for it.
Watch the result that matters
A clearer button might get more clicks but worse inquiries. If your goal is qualified bookings, look at inquiry quality and completed bookings too. Keep the traffic source and offer in mind when comparing weeks.
Maverick's tip
A change in results doesn't automatically prove your edit caused it. Promotions, seasonality, traffic mix and other site changes can move the numbers at the same time.
Save the Research and Check What Changes
Keep the useful findings so you don't restart from scratch next time. Save a summary and the original evidence separately. A summary tells you what you thought; the original screenshots let you check whether that interpretation was fair.
Prompt 14: Save a useful research handoff
Create a concise competitor-research handoff I can save as a text document. Include my business brief, goal, competitors reviewed, date range, evidence index with links or filenames, confirmed findings, interpretations, unanswered questions, chosen changes and the test plan. Keep facts separate from suggestions and clearly label any outdated or incomplete material. Include what we've actually done versus what is still only proposed. Preserve enough context that I can paste this into a new Claude chat without losing the reason behind our decisions. Don't claim you'll automatically remember or monitor anything.
Prompt 15: Tell me what actually changed
Compare this previous research handoff and evidence [PASTE OR UPLOAD] with these fresh competitor pages or captures [URLS OR FILES]. Confirm the dates and whether the materials are comparable before drawing conclusions. Report changes in offers, prices, messaging, ads and customer questions, with before-and-after evidence. Separate a confirmed change from something that was simply absent from our earlier sample. If a page can't be checked, mark it unknown. Explain whether each change affects our chosen strategy. Recommend at most one action worth considering now. Don't invent movement, repeat old findings as new, or create an automated monitoring task. If nothing material changed, say that plainly.
Maverick's tip
Run this again when you have a reason, like a new offer, new ad campaign or a shift in customer questions. You don't have to react to every competitor edit.
Common Mistakes and Quick Fixes
| If this happens | Do this instead |
|---|---|
| Claude praises everything you do | Ask for the strongest case against its recommendation. |
| It only read search snippets | Supply the actual page text or screenshots and rerun that comparison. |
| It calls an old ad a winner | Ask what performance evidence exists. Keep profitability unknown if none does. |
| It gives you generic advice | Ask for the exact page detail, customer concern and proposed change. |
| The competitor is a different kind of business | Compare the closest overlapping offer or choose a more relevant competitor. |
| A simulated buyer sounds certain | Label it a hypothesis and ask a real customer a neutral question. |
| The new copy contains facts you never supplied | Replace them with questions, confirm the facts, then redraft. |
| You've collected more than you can use | Narrow to one business decision and three changes at most. |
Prompt 16: Challenge the analysis before I use it
Audit your competitor analysis as a skeptical reviewer. Identify unsupported claims, stale evidence, unequal comparisons, missing context, speculative customer reactions and any recommendation that copies rather than learns from another business. For each material issue, quote the relevant part of our analysis, explain the problem and rewrite it accurately. Check that evidence links support the actual claim and that numbers use the correct sample and denominator. Then give me the three findings you still have the strongest evidence for, the biggest uncertainty, and the one item I should verify before using the proposed copy. Don't create new certainty to make the report feel finished.
Quick Reference
| Collect | Compare | Improve |
|---|---|---|
| Pages + ads + reviews | The customer decision | One change to test |
| What you want | Start here |
|---|---|
| A guided research plan | Prompt 1: business brief and master instructions |
| Evidence you can trust | Prompt 2: inventory, sources and missing material |
| A clearer website | Prompt 3: compare the buying experience |
| Different customer perspectives | Prompt 4: hypothetical customer decisions |
| Competitor ad research | Prompts 5-6: organize ads and develop original ideas |
| A better ad-to-page connection | Prompt 7: check the promise and the next step |
| A stronger offer | Prompt 8: compare scope, price and conditions |
| Useful customer language | Prompt 9: analyze a balanced review sample |
| Original content ideas | Prompt 10: answer real buying questions |
| A reason to choose you | Prompt 11: specific positioning and proof |
| Draft copy you can use | Prompt 12: page changes and FAQs |
| A practical next step | Prompt 13: seven-day plan and one test |
| A research file you can reuse | Prompts 14-15: save it and compare fresh evidence |
| A final accuracy check | Prompt 16: challenge the conclusions |
Keep these three things
Your original evidence. Your dated research handoff. Your current action plan and results. Keep the old versions so you can see what actually changed.
DO THIS TODAY: Choose one competitor your customer would genuinely consider. Open Claude, paste Prompt 1 and answer its questions. Compare the two websites, pick one useful change, and ask a real customer about the assumption behind it.