The Review page, explained

What the Review page is for

After your AI rewrites finish, Storeshift runs a separate policy-review pass over every rewritten title and body. It scans for terms that have historically triggered TikTok Shop catalog removals — unsubstantiated medical claims, weapon-adjacent terms, unauthorized brand mentions, restricted product categories.

Each match becomes a flag on the Review page. The flag tells you what was matched, what category it falls into, and what TikTok typically does about it. You decide whether to keep, edit, or remove.

The four columns on each flagged row

1. Original eBay text

What the seller (you) originally listed on eBay. This is your factual ground-truth.

2. AI rewrite (current)

The Claude-rewritten version that will go into the CSV. Highlighted terms are the policy matches.

3. Flag(s)

One or more chips like:

• medical_claim — language implying a health benefit ("cures", "treats", "FDA-approved" when not).
• weapon_adjacent — knife / blade / replica / airsoft language even on benign products.
• brand_name — a major brand name (Apple, Nike, etc) that requires authorized-reseller status.
• restricted_category — product types TikTok bans entirely (vaping, certain supplements, adult content).

4. Action buttons

• Keep — accept the rewrite as-is. The flag stays on record but doesn't block the CSV from including this product. Use when you're confident the term is legitimate (e.g. you're an authorized brand reseller).
• Edit — opens the title + body in inline editors. Make any changes you want, save.
• Fix — re-prompts Claude with the specific violation context as steering. The AI rewrites avoiding that exact issue. Free, fast, usually a one-shot fix.
• Remove — drops this product from the CSV entirely. Has an undo button if you change your mind.

How to decide

Keep when:

• You're an authorized reseller of the flagged brand (you have a paper trail for compliance).
• The "claim" is factually true and you can prove it (e.g. "FDA-approved" only if it actually is).
• The flagged term is your distinctive product name (e.g. "Wicked Nun" as the name of your tee design) and you accept the risk that TikTok may reject it.

Fix when:

• You want the rewrite, but with the offending term replaced by something safer.
• You're not sure how to phrase it and want the AI to take another swing with the violation context.

Fix is the right default for most flags. It's free, fast, and the second attempt usually clears the issue while keeping the rewrite quality.

Edit when:

• You want a specific phrasing the AI didn't pick (e.g. you have a preferred way of describing your product line).
• You want to tweak the body HTML manually (re-arrange bullet points, add a sizing note, etc).

Remove when:

• The product itself is restricted (vaping, certain adult products) — TikTok will reject the listing regardless of how you phrase it.
• You don't want to risk it on TikTok but also don't want to make the listing worse just for the channel. You can re-run later without TikTok requirements.
• The product is genuinely off-brand for your Shopify store (one-off old eBay listings you don't want carrying forward).

The "Fix all warnings" button

Top of the Review page is a Fix all warnings bulk button. It runs every flagged row through the regen+apply flow with a concurrency of 3 and a 60-second per-call timeout.

Use it when:

• You have many flags and don't want to handle each individually.
• You trust the AI to make sensible default choices (most flags resolve cleanly).
• You're going to spot-check a few rows after, not every single one.

After it runs, the Review page refreshes with the new rewrites. Anything that's still flagged usually needs a manual decision (Keep / Edit / Remove).

Logo + banner detection

Separate from the policy review, Storeshift's image pipeline detects and drops:

• Repeated store logos appearing in many product photos.
• "Free shipping" stamps and other banner overlays.
• Size charts (which Shopify puts in a separate dedicated field anyway).
• Branded watermarks visible across the catalog.

These are dropped automatically — not surfaced on the Review page — because the heuristics are tuned for high precision (we'd rather miss one logo than drop a product photo). If a real product image was accidentally dropped, edit the product on the Review page and re-attach the original image URL.

Re-exporting after Review

Every action you take (Keep, Edit, Fix, Remove) regenerates the CSV at the same R2 URL. Your original run-complete email link still works after the edit — no need to download a new email. Just re-import the same URL into Shopify when you're ready.

If you've already imported into Shopify before finishing the Review pass, you can re-import the updated CSV with "Overwrite products with matching handles" checked in Shopify Admin. Shopify will update existing products in-place rather than creating duplicates.

What we don't catch

The policy review is heuristic, not exhaustive. TikTok's policies change, and edge cases slip through. Things to also spot-check manually:

• Country-of-origin claims (made-in-USA, etc) — TikTok requires substantiation we can't verify.
• Comparative claims ("better than X", "cheaper than Y") — usually safer to avoid.
• Product category fit — TikTok categorizes from your title; if your title is generic, the wrong category may get assigned and that's a separate problem.

For high-value or high-volume catalogs, a human pass through the top 50-100 products is always worth the 20 minutes.

Questions about a specific flag?

Email hello@storeshift.app with the run ID + product handle. We can look at what triggered the flag and explain the reasoning.