Posting this because it is the wall a lot of people here will hit eventually, and the instinct everyone has is the wrong one. **The thing nobody tells you until you are staring at it** Google's request-review dialog states you get **up to three review attempts**. Not unlimited. Three. That single line should change your whole response to a suspension. If attempts were free, appealing immediately would cost you nothing but time. They are not free. So an appeal submitted before you have changed anything is a wasted third of your budget, checked against the same data that failed. A suspension is an automated flag sitting in a queue. It is not a judgement anyone has made about you. **The timeline, from a store I look after** 05:10 — suspension email. Misrepresentation, whole account, all eight products, UK. Everything dark. Through the day — no appeal. Sweeping the catalogue for anything a policy check could reasonably object to. 19:45 — one review submitted. 00:31 — reinstated. Eight of eight approved, zero not approved. Roughly five hours from submission. First attempt. Two attempts left, unused. **The four things fixed before submitting** 1. **A cosmetic product filed in a medical category.** A hair oil sitting under Hair Loss Treatments. Nobody sets that deliberately — it arrives from an importer, a guess, or a suggestion you accepted months ago. Anti-Aging Skin Care is the same trap: it is a *claim* category, and putting a cosmetic in it is a claim you have not substantiated. 2. **Two products left in vague top-level categories.** If you do not classify precisely, Google classifies for you, and you do not get to see its answer. 3. **A compare-at price of £0.00.** Renders as a struck-through zero next to the real price — an invented discount, squarely inside the same policy. Found while sweeping for anything else that could look misleading. 4. **No contact page and no company details in the footer.** Business identity is part of the same policy surface as product data. It is checkable and it was missing.

This one cost me weeks on a client rebuild in July, and it is invisible from the storefront, so I want it written down somewhere findable. Shopify stores product images in Product Media. That is the system of record. Most themes read it and everything agrees. A heavily customised product template does not have to. It can render its gallery from a metafield, a page builder, or theme asset files loaded through asset_url. When it does, the product page looks exactly as designed and Product Media can be completely out of date without a single visible symptom. Six surfaces show a product image. Your theme decides one of them: 1. Product page — your theme 2. Cart — Product Media 3. Checkout — Product Media 4. Order confirmation email — Product Media 5. Google Shopping feed — Product Media 6. Sales channels (Shop, Meta, TikTok) — Product Media On the rebuild, the product template read image filenames out of a metafield. We shipped forty new product photographs and the site looked superb. The cart, the checkout, the order confirmation email and the Google feed carried on serving the previous owner's photos for weeks, because nothing had ever been written to Product Media. How to check your own store in ten minutes, no code: 1. Add a product to the basket and open the cart. Compare the thumbnail with the product page image. 2. Go through to the checkout and compare again. The checkout is the one you cannot theme your way out of. 3. Open your last order confirmation email and look at the line-item image. 4. In Merchant Center, open a product and look at the image it is serving. 5. In Shopify admin, open the product and look at the Media section. That is what the other five surfaces are reading. If any of those disagree with the product page, your images live in your theme, not in your store. The fix, in the order I would do it: 1. Upload the real image files into Product Media on each product, in the order you want them. 2. Change the theme to read product.media rather than the metafield, so there is one source rather than two.

Most stores assume a product has one name. On Shopify it usually has four, and they can all differ without anything breaking. 1. The heading on the product page. Many themes render this from a metafield or a custom field, not from the product record. 2. The title tag in the HTML. Often written by an SEO app or a keyword sheet, so it reads for search rather than for humans. 3. product.title. The actual product record in Admin. 4. The name printed on the packaging, for physical goods. Why the third one matters more than the other three: product.title is what the cart, the checkout, the order confirmation email and the Google Shopping feed read. Your theme's heading reaches none of those places. So a buyer can pick one product and be asked to pay for a differently named one, and an AI assistant asked for a recommendation can quote a name your own checkout does not use. Nothing 404s, so nothing looks broken, which is exactly why this survives for months. Three checks, ten minutes, no tooling and no apps: Check 1 — view-source the product page. Compare the title tag with the H1 you can see on screen. If an app writes your meta titles, expect a gap. Check 2 — open /products/your-handle.json in a browser tab and read the title field. That is the name in the cart and in the feed. Check 3 — add the product to the basket and read the line item back, then check an old order confirmation email if you have one to hand. The fix is a decision, not a plugin: pick the name that is on the packaging (or the one a customer would say out loud), set product.title to it, and let the heading and the title tag be consistent with it rather than a different name each. Keep keywords in the title tag by all means, but keep the product name inside it recognisable. One caution from doing this on a live store: when you edit SEO fields through the API, send the whole object. Shopify's SEO input nulls any member you leave out, silently, with no error — I wiped the SEO title on eight products that way and only caught it because an unrelated scan flagged it.

During a July client rebuild I ran two live ChatGPT tests: recommend a natural hair oil for the UK. The client's store came up in neither. The store that won had FEWER botanicals - 21 against her 29 - but it named every one and stated the count as a digit. Her pages named six of the 29. The rest sat behind "nature's finest ingredients". That is the whole lesson: an assistant assembles recommendations from what it can quote. Adjectives do not survive retrieval. Named entities, digits and prices do. The fix we shipped: 1. Name everything. Every botanical on the page, including the vernacular names customers actually type - kalonji alongside black seed, methi alongside fenugreek, kadi patta alongside curry leaf. 2. Write counts as digits - "29 botanicals" on the product pages and in llms.txt. A digit is a repeatable fact; "a rich blend" is noise. 3. Put prices in plain text. If a comparing assistant cannot see your price, you are not one of the options. 4. Answer your hardest objection in FAQPage schema. Hers was £49.99 against a £15 supermarket rosemary oil - small batches, cold-pressed, 29 actives against rosemary-plus-carrier. Written down, machine-readable, not buried in a JavaScript accordion. 5. Re-test. Ask the assistant your own category question and read the sentence it quotes about whoever wins. Ten-minute check for your store: open your best product page and ask - if an assistant had to describe this product using only phrases on this page, what would it say? If the answer is adjectives, that is the work. Honest caveat: I cannot promise a rewrite makes the assistant recommend you - there is no submit button. But before the fix the assistant had nothing to quote; now it has 29 named botanicals, a digit, a price and an answered objection. The same rebuild took the store's AI-visibility audit from 71/100 to 100/100: https://www.puniverse.net/case-studies/gold-earth?utm_source=skool&utm_medium=community&utm_campaign=goldearth_case_study_01&utm_content=daily-2026-08-26

Two of our accounts were suspended this year. The Amazon UK seller account was dark for 110 days and came back on the 13th appeal. A Merchant Center account we manage was suspended for Misrepresentation and reinstated the same night, once the field actually feeding Google was fixed. The lesson was not about appeals. It was that every rented channel has an off switch, and we had never written down what would happen if someone flipped it. So now there is a platform-risk ledger. A plain spreadsheet, one row per channel, four columns: 1. Who can switch this off, and for what? Go to the platform's actual policy pages and read the enforcement sections. Summaries and video explainers get this wrong constantly. 2. How would we find out? Amazon told us by email. Google flagged the client account before anyone saw it on a dashboard. Know which alerts you actually have switched on. 3. What is the real appeal timeline? Not the stated turnaround — the realistic one. Ours ran 110 days across 13 appeals. Budget cash and morale for that, not for the five-day promise. 4. How much cash is exposed on day one? Stranded FBA stock, held disbursements, paused ad spend, subscriptions that keep billing while nothing sells. Filling it in takes an hour, and it is an uncomfortable hour: most rows belong to someone else. The two rows you fully control are your own site and your email list, which is the argument for feeding them even while the marketplaces do most of the revenue. What would your worst row look like — which channel, and how much cash sits exposed on it today? Post it below and I will share how we sized ours. If you want the longer version of how we run this company solo, the weekly letter is free, with two guides when you join: https://www.puniverse.net/free?src=skool&utm_source=skool&utm_medium=community&utm_campaign=weekly_letter_evergreen&utm_content=daily-2026-08-23