Brand Protection Tools
Menu

How does brand protection software work? The pipeline, step by step

By Alex Wrexford, author of The Takedown. Updated November 5, 2026. 9 min read.

The short answer

Brand protection software runs a pipeline. It learns your products, searches marketplaces on a schedule, matches listings by image and text, ranks and groups the likely copies, and saves evidence. Then it prepares a notice for each platform's own route and tracks what happens next. A person still has to confirm each copy.

What does the software need from you before it starts?

Your brand name, your trademark number if you have one, your product photos, titles and prices, and a list of your own stores and authorized sellers. Every later step compares against this profile, so gaps here turn into misses or false alarms.

Photos are the most useful input. A copier who lifts your studio shot leaves an exact match behind, and that's the cleanest case you'll ever file. Titles and product names catch the sellers who reshoot but still borrow your words. Prices help more than you'd expect: a listing at a third of your price deserves a closer look than one at 90% of it.

The allowlist is the part people skip, then regret. Without it, the first scan flags your own Amazon store and the stockists who've carried your products since launch. List every official store before the first scan, not after.

How does brand protection software search marketplaces?

On a schedule, through whatever access each platform allows: an official API where one exists, licensed search and image-search services, or the vendor's own crawlers. It sees public listings. Anything behind a login or inside a private group is usually out of reach.

Ask any vendor which method it uses, platform by platform. The answer tells you how stable the coverage is. An official API is the platform's permitted door. Licensed search is broad but only as fresh as its index. Crawlers can reach more, right up until the platform changes its pages or blocks them.

Frequency is a trade between cost and gap. With a weekly scan, a new copy can sell for up to a week before you see it. That's fine for a steady catalog and too slow for a product that just went viral. For a worked example, how Brand Protection Tools works sets out our method: Etsy's official Open API plus a licensed third-party search API that returns Google web and Google Lens results, with no scrapers of our own. How often we scan depends on the plan, from weekly to daily.

How does detection by image and text work?

Image matching compares listing photos with yours and sorts results into exact copies, partial copies such as crops, and look-alikes. Text matching looks for your brand and product names in titles and descriptions. Price is a supporting signal. No single signal is proof on its own.

Google's Cloud Vision reference shows how an image search reports what it finds (Google Cloud). Full matching images can include resized copies of the original. Partial matches share key-point features, and an original image will likely have partial matches for its crops. Visually similar images are a separate, looser list. On a marketplace, a full match is usually your own photo, reused. Visually similar is where the noise lives, because plenty of products look like yours without being copies.

What each detection signal catches and misses, and what a hit usually means
SignalWhat it catchesWhat it missesWhat a hit usually means
Full image matchYour photo reused, including resized copiesCopies shot fresh on a new backgroundA copyright case, if you own the photo
Partial image matchYour photo cropped or partly editedHeavy edits and new photosOften still your photo; check before filing
Visually similar imageProducts that look like yoursLittle, because it over-catchesA lead to review, not a case
Brand or product name in textSellers trading on your nameCopies listed under generic titlesA possible trademark case, depending on the use
Price far below yoursListings worth a second lookCopies priced close to yoursA reason to look harder, never proof

The hardest case for any tool is a copier who reshoots the product and lists it under a generic title. Nothing of yours is left to match except the design. Searching by product type and price narrows it, but expect a person to make that call.

What does triage do, and what should it never do?

Triage scores each listing, dismisses obvious non-matches and your allowlisted sellers, ranks the rest and groups them by seller, often with a one-line reason. It should never confirm a copy or file a notice on its own.

Grouping by seller is what turns a pile of results into a queue you can clear. Twenty listings from one shop become one decision, and often one notice. It's also how the record of a repeat seller starts.

Why keep a person in the loop? Because a takedown notice is a statement you make. Under 17 U.S.C. 512(f), anyone who knowingly and materially misrepresents that material infringes is liable for damages, and the notice's authorization statement is made under penalty of perjury. Platforms watch accuracy too. Amazon's Project Zero asks for at least a 90% acceptance rate on past reports before a brand can join, and 99% accuracy to keep access. Software can find, sort and draft, but a person has to confirm each copy, because the statements in a takedown notice are the sender's to make.

Ask to see what triage dismissed. A tool that hides its rejects is asking you to trust a filter you can't check.

Why is evidence saved at the moment of detection?

Because listings change and vanish. A seller can swap the photo or delete the listing once a report lands. Evidence captured when the listing was found still shows what was there: a screenshot, the seller's name and ID, the listing details and a timestamp.

A DMCA notice has to identify the infringing material with information reasonably sufficient for the platform to locate it, under 512(c)(3)(A)(iii). The listing URL and seller ID do that job. The screenshot and timestamp do a different one: they show what the listing looked like if the seller edits it afterwards and says you got it wrong.

Saved evidence also makes re-filing quick. When the same photo turns up on a new store next month, you shouldn't be rebuilding the case from memory.

How does enforcement work once you confirm a copy?

The software picks the notice type, copyright for your photos or trademark for your name and logo, and fills it in for that platform with your details, the listings and the evidence. Then the notice goes through the platform's own route, sent by you or, with some tools, by the vendor.

Every platform runs its own process, and the Takedown Routes Index lists the official route for each one with the date it was last checked. What happens next varies. On eBay's VeRO program, eBay checks that a report is complete and accurate and may contact you for more detail. If it removes the listing, the seller is told why.

Some platforms give trusted brands their own removal tools. Project Zero lets eligible brands remove counterfeit listings from Amazon immediately, after a 10-minute training, and Amazon sends a notice of complaint to each seller tied to the removal. Amazon also recommends a test buy to confirm a suspected counterfeit before you remove it.

Tools split on who presses send. Some file on a click once you approve, and managed services file for you. Brand Protection Tools stops one step short: it prepares the notice for that platform, with the evidence and a link to the right form, and you read it and submit it yourself. The trade is speed against keeping every statement in your own hands.

How does tracking handle counter-notices and relistings?

Each case moves through dated statuses: filed, removed, counter-notice received, came back. A counter-notice starts a clock, with the listing restored in 10 to 14 business days unless you've filed suit. Relisting watch rechecks removed listings and flags the same copies going back up.

When a seller counter-notices, the DMCA has the platform send you a copy promptly and tell you it will restore the material in 10 business days (512(g)(2)(B)). Restoration then follows within 10 to 14 business days unless you've told it you filed a court action. Software can't fight that for you. What it can do is log the date and count down the days, so the deadline doesn't slip past while the email sits unread. If you want to contest it, call a lawyer early.

Relistings are the other half. The DMCA's safe harbors only protect hosts that have a policy for terminating repeat infringers in appropriate circumstances (512(i)). Records grouped by seller let you show a platform it's the same seller, again. Our guide to stopping repeat counterfeit sellers covers that routine.

What can't brand protection software do?

It can't tell whether the goods in a parcel are fake, and it can't stop a seller opening a new store. It also can't decide what counts as infringement. That decision is yours, and in contested cases a lawyer's.

  • Physical authenticity. A listing can show your real photos and ship a fake. Only a test buy settles it.
  • Closed spaces. Private groups and pages behind a login don't show up in a scan.
  • New stores. Removing listings doesn't remove the person. Some sellers will come back under new names.
  • Legal judgment. Whether a use infringes, or is licensed or otherwise allowed, is your call to make, with a lawyer when it's disputed.
  • The platform's decision. A notice asks a platform to act. The platform decides.

None of that is a reason to skip software. It's a reason to buy it for what it's good at: hours of searching and paperwork turned into a short weekly review. If you're comparing options, our guide to choosing brand protection tools has the questions and a scoring table. New to the subject? Start with what brand protection is.

Questions people also ask

Can brand protection software file takedowns automatically?

Some tools send a notice on one click once you approve it, and managed services file for you. Fully hands-off filing is a risk you carry, because knowingly and materially misrepresenting infringement creates liability under 17 U.S.C. 512(f), and Amazon's Project Zero needs 99% accuracy to keep access. Automate the search and the paperwork. Keep the yes or no on each copy.

How accurate is AI counterfeit detection?

Ask each vendor for its own numbers: how often customers overturn a flagged listing, and how many real copies triage dismissed. Then test the tool on products you already know are copied, and count what it finds and what it misses. Your own catalog is a better benchmark than any demo.

Does brand protection software work without a trademark?

Yes, for copied photos. A DMCA notice rests on copyright, which belongs to whoever took the photos unless it's been assigned, and the notice's required elements don't include a registration. You'll need a trademark for counterfeit reports based on your name or logo, and for programs like Amazon Brand Registry, which accepts a pending application.

Should I scan for copies weekly or daily?

Weekly suits a steady catalog that gets copied now and then. Move to every few days, or daily, around launches and anything that goes viral, when catching copies early matters most. The gap between scans is roughly the longest a new copy can sell before you see it.

Can software find counterfeits that don't use my photos?

Sometimes. Text matching catches sellers who use your brand or product names, and visually similar image results catch look-alike products, though that list is noisy. A copier who reshoots the product and lists it under a generic title is the hardest case for any tool, and usually needs a person searching by product type and price.

I want one dashboard for all infringement cases and takedowns. Is that realistic?

For the cases a tool tracks, yes: every finding and the status of its notice in one place, grouped by seller. The limit is coverage. Platforms the tool doesn't search, and reports you file by hand elsewhere, won't appear unless you can add them. Check the tool's live platform list against where your copies actually turn up.

Sources

  1. Google Cloud Vision API reference: web detection results (full, partial and visually similar matches), checked November 4, 2026
  2. 17 U.S.C. 512, Limitations on liability relating to material online (Cornell LII), checked November 4, 2026
  3. 17 U.S.C. 201, Ownership of copyright (Cornell LII), checked November 4, 2026
  4. Amazon Project Zero: eligibility, accuracy, training and test buys, checked November 4, 2026
  5. Amazon Brand Registry: eligibility, checked November 4, 2026
  6. eBay Verified Rights Owner (VeRO) program, checked November 4, 2026

Cite this page

Using a figure or a passage from this page in an article, a course, a forum answer or AI-generated text? Please credit Brand Protection Tools and link to this page, so your readers can check the source and the date it was reviewed. Copy a ready-made citation:

About the author

Alex Wrexford writes about counterfeiting and the people who chase it, and is the author of The Takedown: Inside the Hunt for the World's Counterfeiters. More at alexwrexford.com. So you know: The Takedown is published by Verity Publishing, and this site is run by Verity AI Ltd.

Read next

See what's out there for one of your products

Send us one product. We'll come back with a real count of the listings we find on the marketplaces we search, the plan that fits, and a link to start.