Infringement searchEvidence gatheringClaim charts

Patent infringement search: finding who uses your invention

By Ipiry TeamPublished August 24, 20267 min read

Short answer

An infringement search runs backwards from a normal patent search: you start with your claims and look for products in the market that practise them. The sources are product listings, technical documentation, regulatory filings and competitors' own patent applications, not patent databases alone. The output is a claim chart with cited evidence, and the search stops being useful the moment it produces assumptions instead of sources.

A patent infringement search runs backwards from the search most people know.

A clearance search starts with your product and looks for patents that might block it.

An infringement search starts with your claims and looks for products in the market that practise them.

Which changes where you look. Patent databases find candidates; they do not prove what a shipping product does. The evidence lives in manuals, specifications, regulatory filings and competitors' own applications.

The two directions

	Clearance / FTO	Infringement search
Starts with	Your product	Your claims
Looks for	Patents that block you	Products that practise them
Primary sources	Patent databases	Product and technical documents
Output	Risk assessment	Claim chart
Consequence	Design change or opinion	Assertion, licence or sale

The source difference is the important one. An FTO search can be done almost entirely in patent databases. An infringement search cannot, because no database records what a product actually contains.

Where the evidence is

Source	Strength
The target's own patent applications	Strongest — self-authored, dated, public
Service and repair manuals	Very strong
Technical specification sheets	Strong
Regulatory submissions	Strong, where public
Published standards compliance	Strong
Teardown reports and testing	Strong
User manuals	Good
Marketing material with technical claims	Moderate
Forum posts, screenshots	Weak
Inference from behaviour	Unusable alone

Competitor patent filings are the most underused source. Companies describe their approach accurately in their own applications because they have to for enablement, and those documents are free and dated.

Service manuals beat user manuals. They describe how the thing works because technicians need to fix it.

Building the candidate list

Step	Method
1	Identify the market segment your claims sit in
2	CPC classification browsing — find the neighbourhood
3	Search competitor filings in those classes
4	Product listings and marketplaces
5	Trade publications and industry directories
6	Standards bodies, where relevant
7	Conference and trade show exhibitor lists

Search by function, in several vocabularies. What you call a flow stabiliser, the market may call a damping baffle or a turbulence damper, and no single term finds everything.

Classification browsing catches what keywords miss, because CPC groups things by what they do rather than what they are called.

Screening candidates

Test	Action
Does the product exist in the relevant market?	Rights are national
Is it currently sold?	Past sales still matter within 6 years
Does public documentation exist?	Determines whether it is provable
Does an independent claim element appear absent?	Stop here
Volume and revenue	Determines whether pursuit is proportionate

Screen before mapping. Most candidates fail on jurisdiction, volume, or an obviously absent element, and screening them out cheaply is the point.

A missing independent claim element ends it. Falling outside an independent claim means falling outside every claim that depends on it.

Mapping what survives

Column	Contains
Element	The exact claim words, one per row
Accused feature	The specific corresponding feature
Evidence	Source with a page or figure reference
Grade	Present / absent / arguable / unverified
Date of source	Products change

One row per element, splitting at every limitation. A phrase containing two requirements is two rows.

Grade "unverified" rather than assuming. An internal component you cannot observe is unverified, and saying so tells the reader exactly what evidence would settle it.

Date every source. A 2022 service manual may not describe the 2026 model, and that gap gets found by the other side if not by you. See product mapping patent infringement.

Method claims add a question

Extra step	Detail
Identify who performs each step	Add a performer column
One party performs all	Direct infringement
Steps split across parties	Divided infringement
Direction or control shown?	Required to attribute

Divided infringement is a real defence where a service performs some steps and a customer performs others.

Map the actor for every step, then ask whether one party directs or controls the rest. An ordinary vendor-customer relationship generally does not.

What the search cannot tell you

Question	Why not
Whether the claims are valid	A different search entirely
Whether estoppel bars equivalents	Prosecution history, not the market
What a court will construe the terms to mean	Judgement
Internal implementation details	Frequently unobservable
Whether pursuit is worth it	Commercial decision

An infringement search is one input. A strong chart against a product covered by an invalid patent is worth nothing, and a gap bridged by equivalents that estoppel bars is worth nothing either.

Check the prosecution history before relying on equivalents. It is free, public, and frequently decisive. See prosecution history estoppel.

Worked example: from claims to a chart

A patent on a filtration control method. Six independent claim elements.

Stage	Action	Result
1	CPC browsing in the filtration control classes	40 competitor filings
2	Filtered to US-market products	12 companies
3	Screened on public documentation	7 with usable material
4	Screened on an obviously absent element	3 survive
5	Mapped from service manuals and their own filings	—

The chart on the strongest candidate

Element	Evidence	Grade
1. Fluid inlet	Service manual, fig. 3	Present
2. Sensor upstream	Spec sheet, p.2	Present
3. Controller	Their own application, para. 41	Present
4. Threshold comparison	Their application, para. 44	Present
5. Valve actuation on exceedance	Service manual, §5.2	Present
6. Feedback loop with damping	Not documented publicly	Unverified

Five of six elements proven from public documents. Two of those came from the competitor's own patent application.

The sixth requires a teardown or discovery, and the chart says so rather than assuming it.

That chart is what makes the patent an asset. A patent with documented evidence against a shipping product is a different thing from one without.

Before sending anything

Consideration	Effect
A notice letter starts the knowledge clock	Supports later willfulness
It can also trigger a declaratory judgment suit	In their chosen forum
Marking starts damages without a letter	Avoids the trade-off
Validity risk	A patent asserted is a patent challenged
Six-year damages cap	Delay costs the earliest years

Marking is the option that starts damages without inviting a filing. See patent marking.

Assertion puts the patent itself at stake. An IPR filed in response can end the asset, not just the claim.

The search before a fee decision

Ipiry Patent Survival Curve v1.0	Rate
Survive the 3.5-year fee (2022 cohort)	85.8%
Survive the 7.5-year fee (2018 cohort)	64.6%
Reach full term (2014 cohort)	41.4%
Abandoned before full term	58.6%

Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents — see the patent survival curve.

Whether anyone practises the claims is the question that decides the fee. Running a screening search before each window converts a payment into a decision.

It also determines whether selling is realistic. A documented chart is the largest single factor in what a buyer will pay.

Where automation helps

Stage	Automation
Candidate identification	Substantial
Classification browsing	Substantial
Document retrieval	Substantial
First-draft mapping	Substantial
Verifying every citation	Human
Deciding what to pursue	Human

Every cited source needs checking. A fabricated or misread citation is worse than an empty row, because it makes the whole chart untrustworthy.

Patent infringement search: the checklist

Start from your claims, not from your product.
Use CPC classification browsing to find the neighbourhood keywords miss.
Search competitors' own patent filings. They describe their approach accurately.
Prefer service manuals to user manuals, and both to marketing material.
Screen on jurisdiction and an obviously absent element before mapping.
Map one element per row with a locatable, dated source.
Grade honestly. "Unverified" is more useful than an assumption.
Add a performer column for method claims and consider divided infringement.
Check the prosecution history before relying on equivalents.
Run a screening search before every fee decision, and before any sale.

Common questions

What is a patent infringement search?

A search for products or processes in the market that practise the claims of a patent you own. It runs in the opposite direction from a clearance search, which starts with your product and looks for patents that might block it.

Where do you look for infringers?

Product listings, technical specifications, user manuals, service documentation, regulatory filings, published standards, teardown reports, and competitors' own patent applications. Patent databases are useful for finding candidates, not for proving what a product does.

Why are competitor patent filings a good source?

Because companies describe their own technical approach accurately in their applications, and those documents are public. A competitor's application describing an approach that reads on your claims is strong, dated, self-authored evidence.

How do you search when you don't know the vocabulary?

Search by function rather than by your own name for the thing, in several vocabularies, and use CPC classification browsing to find the neighbourhood. What you call a flow stabiliser the market may call a damping baffle.

What if the infringing feature is internal?

Then it may be unverifiable from public sources. Grade it as unverified rather than assuming it, and note what evidence would settle it — a teardown, a service manual, or a regulatory submission. An honest gap is more useful than a guess.

Does finding a similar product mean infringement?

No. Infringement requires every element of at least one claim to be present. Overall similarity is not the test, and a product that looks alike may omit a claimed element while a product that looks different may include them all.

How much does an infringement search cost?

The initial screening can be done for nothing using free tools and public product information. Detailed verification, teardowns and professional analysis cost more, and are worth reserving for candidates that survive screening.

What do I do with the results?

Build a claim chart mapping each element to cited evidence, then decide whether to pursue. Sending a notice letter has consequences, including starting a knowledge clock and potentially inviting a declaratory judgment suit in a forum you did not choose.

Should I search before selling a patent?

Yes. Evidence that a patent reads on products in the market is the single largest factor in what a buyer will pay. A patent with a documented claim chart against a shipping product is a different asset from one without.

Can the search be automated?

Candidate identification benefits substantially from automation, and first-draft charts save real time. Every cited source still needs verification, because a fabricated or misread citation is worse than an empty row.

How often should I run one?

Periodically, and certainly before any maintenance fee decision. Whether anyone practises the claims is the question that decides whether a fee is worth paying, and it changes as markets develop.

What is the difference from infringement detection?

Detection usually refers to ongoing monitoring that surfaces new candidates automatically. A search is the deliberate exercise of looking, at a point in time, and the two work together — monitoring flags, searching verifies.

Patent monitoring: what to watch and why

What patent monitoring covers, the five things worth watching, how alert thresholds work, what a monitoring programme costs in attention, and how to act on what it finds.

Product mapping methodology for patent infringement

A repeatable method for mapping a product against patent claims — parsing the claim, gathering evidence, building the chart, grading each element, and knowing when the mapping is finished.

Patent claim chart: how to build one

A claim chart maps every element of a patent claim to a feature of an accused product or a prior art reference. How to build one, what makes it credible, and where they are used.

What is your patent worth?

Free AI valuation in 60 seconds, benchmarked against 652,000 KPSS patent value records and the Ipiry Patent Survival Curve.

Value My Patent Free →Takes 60 seconds. No credit card required.