Turn your website into a week of ads with Claude

Read this page and build me a brand brief I can run paid social and search ads from: [paste your URL].

Tell me back, in plain language: what the product or service actually is, who is clearly being sold to and who is not, the three strongest pieces of proof on the page (reviews, numbers, guarantees, named credentials, client logos), the price and how it is framed, and the single biggest objection a first-time buyer would have. Then describe the brand voice in five specific adjectives with a real quoted line from the page for each one, so I can tell whether you actually read it or guessed.

Flag anything you are inferring rather than reading directly off the page, because I am not putting an unverifiable claim into a live ad. If the page is thin on proof or vague about who it is for, say so directly instead of writing around the gap. Do not use em dashes.

Using the brand brief above, write me ten paid social ad variants built on ten genuinely different angles. An angle is a reason to care, not a rephrasing, so make them distinct from each other: include at least one problem-first, one proof-first, one comparison, one identity, one objection-handling, and one that leads with the specific price or offer.

For each of the ten give me: the angle name, a hook line under twelve words, primary text between 50 and 150 characters, and a headline under 27 characters. Write to the platform's recommended lengths, not over them, because copy that gets truncated is copy you did not really write. Keep the language plain and specific, use the voice adjectives from the brief, and never invent a statistic, guarantee or testimonial that is not on the page.

Then give me the whole thing as a table with columns for variant number, angle, hook, primary text, headline and character counts, so I can paste it straight into a build sheet. At the end, tell me which two angles you think are weakest and why, and I will decide whether to cut them.

Write one email that continues the strongest angle from the ten variants, for someone who clicked that ad and gave me their email but has not bought yet.

Give me three subject lines under 45 characters, one preview line, and a body between 120 and 200 words. Open on the same tension the ad opened on rather than restating the offer, use one specific piece of proof from the brand brief, handle the main objection you identified, and finish with a single clear call to action. One call to action, not three. Match the voice adjectives from the brief and keep the reading level plain.

Then write me a two-line plain-text version for people who read email on their phone in ten seconds, and tell me which of the three subject lines you would send first and why. No em dashes anywhere.

Write image briefs for the top four variants from the table, detailed enough that a designer who has never seen my product could build them in Canva without asking me a follow-up question.

For each brief give me: the concept in one sentence, the exact on-image copy with a character count, the visual composition (what is in frame, where the product sits, what the background is, whether a person appears and what they are doing), the mood in three words, and where the negative space sits so the headline has somewhere to live. Build every brief at a 4:5 ratio, 1440 by 1800 pixels, which is what the platform's own ads guide recommends for feed image ads.

Then add a "do not do" line to each brief listing the two most likely ways a designer could misread it. Finish with a file naming convention in the pattern ANGLE_VARIANT_DATE, and write out the four filenames, because in three weeks those names are the only thing connecting my performance data to what I actually made.

Build me a two-week testing plan for these variants, assuming a total budget of [your budget] and that I am testing for [your goal: leads, purchases, or bookings].

Week one: tell me which variants run first and why, how to group them so the comparison is readable, and the one variable each pairing is testing. Hold everything else constant, and say explicitly what I must not change mid-test. Give me the minimum I should spend before I am allowed to judge anything, and be honest that at a small budget some of these results will never reach a confident answer.

Week two: tell me what to do with the winner, what to do with a variant that is cheap but low quality, and what to kill outright. Give me a plain decision rule for each, in the form "if X then do Y", so I am not making the call on vibes at 11pm. Finish with the three numbers I should be looking at for my goal, in priority order, and one sentence on what each one actually tells me.