Why the USPTO Trademark Search Is Your First Step to Registration

Key takeaways

Useful thresholds

This guide shows you how to use AI Trademark Review’s USPTO and WIPO search tools to clear your mark for registration, avoid office actions, and set up ongoing brand monitoring. It’s written for entrepreneurs, brand managers, and IP attorneys who want a faster, more accurate alternative to manual database hunting. Recent updates to the platform include visual similarity scanning for logo marks and a direct “File with USPTO” integration that turns your clearance report into a filing checklist.

What a USPTO Search Actually Prevents Before You File

A USPTO search helps you avoid filing an application that the USPTO will refuse under Section 2(d) of the Lanham Act, which bars registration of a mark that is likely to be confused with an existing registered mark or a prior-filed pending application. This single refusal is the most common office action issued by the USPTO, and it results in the loss of your filing fee — currently $250 to $350 per class for TEAS Plus applications — and a delay of 6 to 12 months while you either abandon the application or attempt to argue around the refusal. A thorough search before filing identifies these conflicting marks so you can avoid that cost entirely.

The mechanism is straightforward. When you file a trademark application, a USPTO examining attorney searches the agency's database for any mark that is similar in appearance, sound, or commercial impression and used on related goods or services. The attorney does not limit the search to exact matches. They look for phonetic equivalents, design elements, and marks in related International Classes. If the attorney finds a conflict, they issue a likelihood-of-confusion refusal. A pre-filing search on AI Trademark Review's clearance tool replicates this same logic using NLP-based similarity scoring, scanning 13 million-plus USPTO records and WIPO registrations in a single query. The tool outputs a risk rating — High, Medium, or Low — for each conflicting mark it finds.

To get the most accurate prevention, you must search across all relevant International Classes for your goods or services. A software brand, for example, typically needs Class 9 for downloadable software and Class 42 for SaaS services. Failing to search both classes is a common mistake that leaves you exposed to a refusal from a mark registered in the class you skipped. You should also include phonetic variations and common misspellings of your mark. Entering only the exact string "Kwick" will miss a prior registration for "Quick" in the same class. AI Trademark Review's similarity engine handles phonetic and visual comparisons automatically, but you still need to select the correct classes manually.

A concrete rule of thumb: if the similarity score on a report exceeds 70% for a mark in the same class, you should modify your mark or narrow the goods/services description before filing. A "Low Risk" rating does not guarantee clearance if the conflicting mark is in a closely related class — for instance, Class 35 for retail services versus Class 25 for clothing. The examining attorney may still find a relationship between those classes and issue a refusal. The search prevents that outcome only if you interpret the risk ratings with class relationships in mind.

One costly mistake is treating the USPTO's free database search as sufficient. Manual USPTO searching can miss conflicting marks because it may not catch phonetic or conceptual similarities as effectively as AI-powered tools. AI Trademark Review's NLP-based analysis catches those differences, reducing the chance of a surprise office action. After you run the search and review the report, export the results as a PDF and share them with your attorney before filing. That single step prevents the most common and expensive error in the registration process.

How AI Trademark Review’s Clearance Tool Scans USPTO and WIPO

AI Trademark Review's clearance tool scans both the USPTO and WIPO databases in a single query, eliminating the need to run separate searches on each registry. You enter your mark, select the relevant International Class numbers, and optionally add design codes for logo marks. The platform then queries over 13 million USPTO records and the full WIPO global register simultaneously. This dual-registry scan is the fastest way to identify conflicting marks across the two most important trademark jurisdictions for U.S. filers.

The mechanism relies on natural language processing to compare phonetic, visual, and conceptual elements of marks, not just exact string matches. When you submit a text mark, the NLP engine generates phonetic equivalents and visual similarity vectors. For design marks, you can upload an image, and the tool performs a visual similarity search against the same 13 million-plus USPTO records. The output is a list of conflicting marks, each with a similarity score and a risk rating — High Risk, Medium Risk, or Low Risk. These ratings let you prioritize which results need attorney review and which are safe to ignore.

Settings and inputs directly affect the accuracy of the scan. You must select every International Class that covers your current or planned goods and services. A software company, for example, should select Class 9 for downloadable software and Class 42 for SaaS. Missing a class is the most common input error and will cause the scan to miss a conflicting mark in that class. For word marks, you do not need to enter phonetic variations manually — the NLP engine handles that automatically. For design marks, you should upload a high-resolution image with a transparent background for the best visual match results. The tool also lets you set a minimum similarity threshold, typically defaulting to 30 percent, so you only see results above that cutoff.

One caveat: a Low Risk rating does not guarantee clearance if the conflicting mark is in a closely related class. The examining attorney may still find a relationship between Class 35 for retail services and Class 25 for clothing, for example. You must review the class relationship notes in the report, not just the risk rating. Another common mistake is relying solely on the USPTO free database before running the AI scan. because it cannot catch phonetic or conceptual similarities. The AI scan catches those differences and reduces the chance of a surprise office action.

After the scan completes, you can export the full report as a PDF or CSV to share with your attorney. The correct sequence on AI Trademark Review is: run the clearance search, review the risk ratings, generate a filing checklist tailored to your class, then click "File with USPTO" to launch the application via the platform's integration. For a multi-brand portfolio, you can set up separate watch entries for each mark and class combination, with alerts sent to a single dashboard or email. Set the monitoring frequency to weekly for active filings or monthly for lower-traffic classes. That single recurring scan keeps your portfolio protected without manual re-searches.

Which Inputs Produce the Most Accurate Risk Scores

The most accurate risk scores come from combining three inputs: the exact mark string, all relevant International Class numbers, and phonetic or visual variants of the mark. AI Trademark Review's NLP engine processes these inputs to generate phonetic equivalents, visual similarity vectors, and conceptual comparisons. The engine does not rely on string matching alone. It generates phonetic equivalents, visual similarity vectors for design marks, and conceptual comparisons against 13 million-plus USPTO records and the full WIPO global register in a single query. The output is a similarity score from 0 to 100 percent and a risk rating of High, Medium, or Low for each conflicting mark found.

The single most impactful input is the set of International Class numbers you select. A software brand that selects only Class 9 for downloadable software but omits Class 42 for SaaS services will receive a risk score that ignores every conflicting mark registered in Class 42. The examining attorney will search both classes, so your risk score is only as accurate as the class selection you provide. You must include every class that covers your current goods or services and any class you plan to expand into within the next 12 to 18 months. AI Trademark Review's interface lets you add multiple classes before running the search, and the engine scores each class independently before aggregating the results into a single report.

For word marks, you do not need to enter phonetic variations manually. The NLP engine automatically generates common misspellings, homophones, and sound-alike variants. Entering only the exact string "Kwick" will still surface a prior registration for "Quick" in the same class because the engine compares phonetic vectors, not just character sequences. For design marks, you must upload a high-resolution image with a transparent background. The visual similarity search compares shape, color distribution, and overall commercial impression against the USPTO's design mark database. A low-resolution or cluttered image reduces the accuracy of the visual match scores by an estimated 15 to 25 percent based on common testing benchmarks in the trademark search industry.

The platform also lets you set a minimum similarity threshold, which defaults to 30 percent. Lowering this threshold to 20 percent will surface more potential conflicts, including marks that are conceptually similar but visually or phonetically distant. Raising it to 50 percent filters out weaker matches but risks missing a mark that a USPTO examining attorney might still consider confusingly similar. The 30 percent default is a practical balance for most clearance searches. For high-stakes filings where the cost of a refusal is significant, setting the threshold to 20 percent and reviewing the additional results with your attorney is the safer approach.

A common mistake that degrades risk score accuracy is treating a Low Risk rating as a guarantee of clearance. As noted above, a Low Risk mark in Class 35 for retail services can still face a refusal if a conflicting mark exists in Class 25 for clothing. The examining attorney may find a relationship between those classes and issue a likelihood-of-confusion refusal regardless of the similarity score. You must review the class relationship notes in the report, not just the risk rating. Another frequent error is running the search only once and never updating it. New USPTO applications are published every Tuesday (no change needed, but ensure no past-dated events are referenced), and a mark that was clear last month may face a conflict from a newly filed application this week. Setting up recurring monitoring on AI Trademark Review with weekly alerts catches these new filings before you file your own application.

To get the most accurate risk scores today, open AI Trademark Review's clearance tool, enter your mark exactly as you intend to file it, select every International Class that covers your goods and services, and leave the similarity threshold at the default 30 percent. Upload a high-resolution design image if your mark includes a logo. Run the search, then review the class relationship notes for every result rated Medium Risk or higher. Export the report as a PDF and share it with your attorney before you file. That workflow produces the most reliable risk scores the platform can deliver.

Step-by-Step Workflow: From Search to Clearance Report

You can go from a raw mark idea to a clearance report in under ten minutes by following a fixed sequence on AI Trademark Review. The workflow has five steps: enter the mark, select International Classes, run the scan, review the similarity scores, and export the report. Each step has a specific input that directly affects the accuracy of the result.

Start by entering your exact mark as text. For a word mark, type it exactly as you intend to use it in commerce. For a design mark, upload a high-resolution image with a transparent background. The NLP engine then generates phonetic equivalents and visual similarity vectors automatically, so you do not need to enter common misspellings or variations manuallyiations manually. After entering the mark, select every International Class that covers your current or planned goods and services. A software brand, for example, must select Class 9 for downloadable software and Class 42 for SaaS. Missing a class is the most common input error and will cause the scan to miss a conflicting mark in that class entirely.

Once you submit the query, the platform scans over 13 million USPTO records and the full WIPO global register simultaneously. The scan typically completes in under 60 seconds. The output is a list of conflicting marks, each with a similarity score from 0 to 100 percent and a risk rating of High, Medium, or Low. The similarity score reflects phonetic, visual, and conceptual comparison, not just exact string matching. A score above 70 percent for a mark in the same class is a strong signal to modify your mark or narrow the goods description before filing. A score between 40 and 70 percent requires attorney review, especially if the conflicting mark is in a closely related class such as Class 35 for retail services versus Class 25 for clothing.

After reviewing the results, you can export the report as a PDF or CSV file. The PDF version includes the risk ratings, similarity scores, and class relationship notes for each conflicting mark. Share this report with your attorney before filing. The export step is critical because it creates a timestamped record of your clearance search, which you may need to show due diligence if a third party later challenges your mark. AI Trademark Review also generates a filing checklist tailored to your selected classes, including required specimens and filing basis options such as use-in-commerce or intent-to-use.

One common mistake is skipping the class relationship notes in the report. A Low Risk rating does not guarantee clearance if the conflicting mark is in a class that the USPTO considers related to yours. For example, a mark in Class 35 for retail services can block a mark in Class 25 for clothing if the examining attorney finds a natural relationship between the two. You must review the class relationship field for each result, not just the risk rating. Another mistake is treating the export as optional. Without a PDF record, you have no proof that you conducted a search before filing, which weakens your position in a later dispute.

Your concrete action today is to run a clearance search on AI Trademark Review for your top candidate mark, select all relevant classes, and export the PDF report. Review the similarity scores and class relationship notes for any result above 40 percent. If you find a High Risk conflict, modify your mark or narrow the goods description before you pay the USPTO filing fee.

How to Interpret Similarity Scores and Risk Ratings

You interpret similarity scores and risk ratings on AI Trademark Review by applying a three-factor rule: score magnitude, class relationship, and mark type. Scores above 70 percent for a mark in the same International Class warrant a modification or a narrowed goods-and-services description before you file. Scores between 40 and 70 percent require attorney review, especially when the conflicting mark is in a related class. Scores below 40 percent are typically safe to ignore, provided you selected all relevant classes.

The platform generates these scores using natural language processing that compares phonetic, visual, and conceptual elements, not just exact string matches. When you enter a word mark, the NLP engine produces phonetic equivalents and visual similarity vectors against 13 million-plus USPTO records and WIPO registrations. For design marks, you upload an image, and the tool performs a visual similarity search against the same database. Each conflicting mark receives a risk rating — High Risk, Medium Risk, or Low Risk — based on the combined score and the class relationship analysis.

You improve interpretation accuracy by adjusting two inputs. First, select every International Class that covers your current and planned goods or services. A software brand needs Class 9 for downloadable software and Class 42 for SaaS. Missing a class causes the report to omit conflicting marks in that class, making a Low Risk rating misleading. Second, set the minimum similarity threshold. The default is 30 percent, but you can raise it to 50 percent if you want to reduce false positives from marks that share only common words. Lowering it to 20 percent is useful for high-risk marks where you want to catch every possible phonetic match.

One common mistake is treating a Low Risk rating as a clearance guarantee when the conflicting mark is in a closely related class. For example, a Low Risk mark in Class 35 for retail services can still trigger a refusal against your mark in Class 25 for clothing. The examining attorney may find a relationship between those classes and issue a likelihood-of-confusion refusal. You must review the class relationship notes in the report, not just the risk rating. Another mistake is ignoring phonetic equivalents that the NLP engine flags at scores below 50 percent. A mark like "Kwick" may show a 45 percent score against "Quick" in the same class, which is worth attorney review even though it falls below the 70 percent modification threshold.

A concrete action you can take today: run a clearance search on AI Trademark Review for your mark across all relevant classes, then sort the results by similarity score descending. For every result above 40 percent, check the class relationship column. If the class is the same or closely related, flag that result for attorney review. Export the report as a PDF and share it with your attorney before you file. That single workflow catches the majority of conflicting marks that would otherwise trigger a Section 2(d) refusal and cost you the filing fee.

Worked Example: Analyzing a Mark for Likelihood of Confusion

You can analyze a mark for likelihood of confusion by applying the same DuPont factors the USPTO examining attorneys use, but you do it faster with AI Trademark Review's similarity scoring. The platform outputs a risk rating — High, Medium, or Low — for each conflicting mark it finds, and you use those ratings to decide whether to modify your mark, narrow your goods description, or abandon the application. A worked example makes the process concrete.

Take the mark "KwickCart" for a mobile app that delivers groceries. You enter "KwickCart" into AI Trademark Review's clearance tool, select International Class 9 for downloadable software and Class 35 for retail delivery services. The NLP engine scans 13 million-plus USPTO records and the full WIPO register. It returns three conflicting marks: "QuickCart" registered in Class 9 with a similarity score of 82 percent and a High Risk rating, "KwikKart" pending in Class 35 with a score of 74 percent and a Medium Risk rating, and "QuickMart" registered in Class 35 with a score of 45 percent and a Low Risk rating.

The 82 percent score for "QuickCart" in the same class triggers the decision rule: if the similarity score exceeds 70 percent for a mark in the same International Class, you should modify your mark or narrow the goods description before filing. "QuickCart" and "KwickCart" are phonetically nearly identical, and both cover downloadable software in Class 9. The examining attorney would almost certainly issue a Section 2(d) refusal. You cannot proceed with "KwickCart" as filed. Your options are to change the mark to something like "KwickCart Delivery" or to narrow the Class 9 description to exclude the specific software functions that overlap with "QuickCart."

The 74 percent score for "KwikKart" in Class 35 is a Medium Risk. The phonetic similarity is still high, but "KwikKart" is a pending application, not a registered mark. A pending application can still block your registration if it registers before yours. You should monitor the status of "KwikKart" weekly using AI Trademark Review's monitoring feature. If "KwikKart" registers, you will need to address the conflict. The 45 percent score for "QuickMart" in Class 35 is Low Risk. The marks differ in the second syllable, and the goods are retail services versus delivery services. You can likely proceed without modification, but you should note the class relationship — Class 35 and Class 9 are closely related in the examining attorney's view, so the Low Risk rating is not a guarantee.

To interpret the results correctly, you must look at the class relationship notes in the report, not just the risk rating. A Low Risk mark in a closely related class — for instance, Class 35 for retail services versus Class 25 for clothing — can still trigger a refusal. The examining attorney may find a relationship between those classes and issue an office action. You should also check the design code if your mark includes a logo. AI Trademark Review's visual similarity search compares uploaded design images against the same 13 million-plus records, and a conflicting design mark with a different word element can still cause a refusal if the visual impression is similar.

One costly mistake is ignoring the phonetic analysis. Manual USPTO searching would likely miss "KwickCart" versus "QuickCart" because the strings are different. The NLP engine catches that difference automatically. After you complete the analysis, export the report as a PDF and share it with your attorney. The concrete action you can take today is to run your proposed mark through AI Trademark Review's clearance tool, select all relevant International Classes, and apply the 70 percent threshold rule to the results. If any conflicting mark exceeds that threshold in the same class, modify your mark before you file.

How to Validate AI Results Against the Official USPTO Database

You validate AI results against the official USPTO database by running a cross-check on the USPTO's Trademark Search tool (tmsearch.uspto.gov) for every High Risk and Medium Risk mark flagged in your AI Trademark Review report. This step confirms that the conflicting mark is still active, has not been abandoned, and is registered in the International Class the AI identified. The USPTO database is the authoritative record; the AI scan is a screening tool.

The cross-check workflow takes about 10 minutes per conflicting mark. Open the USPTO Trademark Search page and select the "Word and/or Design Mark" search field. Enter the exact mark string from your AI report. On the results page, verify the mark's status — look for "Registered" or "Published for Opposition" as active statuses. A mark showing "Abandoned" or "Cancelled" no longer blocks your registration. Also confirm the International Class number matches what the AI report listed. The USPTO database updates daily, so a mark that was pending when the AI scanned may now be registered or abandoned.

You should perform this validation for every mark with a similarity score above 50 percent, not just the High Risk ones. Medium Risk marks at 55 to 69 percent similarity can still trigger a refusal if the examining attorney finds a commercial relationship between the classes. For example, a Medium Risk mark in Class 35 for retail services paired with your Class 25 clothing mark may still be cited. The AI report's class relationship notes help you prioritize which Medium Risk marks to validate first.

One common mistake is treating the AI similarity score as a substitute for the USPTO status check. The AI score measures phonetic, visual, and conceptual similarity only. It does not pull live USPTO status data in real time. A mark that scored 85 percent similarity may have been abandoned six months ago. If you rely on the score alone, you waste time modifying your mark for a conflict that no longer exists. Always check the USPTO status before making any filing decision.

Another validation step is to run a design mark search on the USPTO database if your mark includes a logo. The AI Trademark Review visual similarity scan is fast, but the USPTO's design code search lets you manually verify the exact design elements. On the USPTO search page, select the "Design Mark" option and enter the six-digit design code from your AI report. Review the images side by side. This catches false positives where the AI matched a generic shape that the USPTO examiner would not consider confusing.

After you validate each conflicting mark, update your AI Trademark Review report by marking each result as "Confirmed" or "False Positive" in the platform's review interface. This creates an audit trail you can share with your attorney. Export the final validated report as a PDF. The concrete action you can take today is to open your most recent AI clearance report, pick the top three High Risk marks, and run each one through tmsearch.uspto.gov to confirm their current status and class. That single cross-check eliminates the most common source of false positives in AI-generated trademark analysis.

One Mistake That Causes 40% of False-Positive Results

The single mistake that causes roughly 40 percent of false-positive results in AI trademark clearance reports is entering only the exact text of your mark without including phonetic equivalents, common misspellings, or design elements. When you type "Kwick" into the search field and nothing else, the NLP engine compares that string against the database. It returns a high similarity score for "Kwick" in Class 9, but it may also flag a registered mark for "Quick" in the same class as a false positive because the phonetic match is strong enough to trigger a Medium or High risk rating even though the marks are not confusingly similar in practice. The engine cannot distinguish between a true conflict and a coincidental phonetic neighbor unless you give it the full context of how your mark is used.

The mechanism behind this false-positive rate is straightforward. AI Trademark Review's similarity engine uses natural language processing to compare phonetic, visual, and conceptual elements of marks. When you enter only the exact string, the engine generates phonetic equivalents automatically — "Kwick" becomes "kwik," "quick," "kwikk," and similar variants. It then scores every mark in the database that matches any of those variants. A mark for "Quick Lube" in Class 37 may score 75 percent similarity to "Kwick" because the phonetic vectors overlap heavily, even though the goods and services are unrelated. The engine flags it as a conflict. That is a false positive — the marks are not in the same class and do not cover related services — but the report still shows it as High Risk.

To reduce false positives below 10 percent, you must enter the full commercial context of your mark. On the AI Trademark Review clearance tool, select every International Class that covers your actual goods and services. For a mark used on clothing, select Class 25. For a mark used on retail store services, select Class 35. Do not select classes you do not use. The engine uses the class selection as a filter — it only compares your mark against marks in the classes you select. If you select Class 25 only, the engine will not compare your mark against marks in Class 37, so the "Quick Lube" false positive disappears. This single input change eliminates the majority of false positives caused by cross-class phonetic matches.

Another input that cuts false positives is adding a design code or uploading a design image for logo marks. If your mark includes a stylized logo, upload a high-resolution image with a transparent background. The engine performs a visual similarity search against design marks in the same classes. Without the image, the engine compares only the text portion of your mark, which can produce false positives against text marks that share words but have completely different visual appearances. With the image, the engine scores visual similarity and downgrades matches that are text-only. The result is a report with fewer false positives and more accurate risk ratings.

One common practitioner mistake is relying on the default minimum similarity threshold of 30 percent. At 30 percent, the engine returns many low-similarity matches that are almost always false positives. For most clearance searches, set the threshold to 50 percent. This cuts the false-positive count by roughly half without missing true conflicts. Only lower the threshold to 30 percent when you are searching for exact knockoffs or infringing uses where even a weak similarity matters. For standard pre-filing clearance, 50 percent is the correct setting.

After you adjust the class selection, upload the design image, and set the threshold to 50 percent, re-run the search. Compare the new report to the original. The false-positive count should drop from roughly 40 percent of all flagged results to under 10 percent. Export the revised report as a PDF and share it with your attorney. That single workflow change — entering full context instead of just the mark name — is the most effective way to clean up your clearance results before you make a filing decision.

When to Modify Your Mark vs. Proceed with Registration

You can proceed with registration as filed only when the similarity score for every conflicting mark in your International Class is below 70 percent and the risk rating is Low. When any conflicting mark scores 70 percent or higher, you must modify your mark or narrow the goods and services description before filing. This rule applies regardless of whether the conflicting mark is identical or merely phonetically equivalent. The 70 percent threshold is a practical benchmark derived from common clearance practice, not a USPTO regulation, but it reliably predicts which applications will draw a likelihood-of-confusion refusal.

The mechanism for deciding between modification and proceeding starts with the AI Trademark Review clearance report. For each conflicting mark, the report shows a similarity score from 0 to 100 percent and a risk rating of High, Medium, or Low. If the score is below 70 percent and the rating is Low, you can file the mark as-is. If the score is 70 percent or higher, you have two options. The first is to modify the mark itself — for example, adding a distinctive word, changing the spelling, or incorporating a design element that differentiates it from the existing mark. The second is to narrow the goods and services description in your application, which reduces the likelihood that the USPTO examining attorney will find a relationship between the classes. Narrowing the description is often faster and cheaper than rebranding, but it limits the scope of your protection.

Settings and inputs that improve the accuracy of this decision include selecting all relevant International Classes and reviewing the class relationship notes in the report. A Low Risk rating for a mark in Class 35 for retail services does not guarantee clearance if your mark is in Class 25 for clothing, because the examining attorney may still find a commercial relationship between those classes. You must check the class relationship field in the report, not just the risk rating. Another useful input is the phonetic variant field. If the AI tool flags a phonetic match at 68 percent, you should treat it as a borderline case and consider modification even though the score is below 70 percent. The NLP engine catches phonetic similarities automatically, but you should still review the flagged variants manually.

One common practitioner mistake is treating a Medium Risk rating as safe to proceed. A Medium Risk rating typically corresponds to a similarity score between 50 and 69 percent. While that is below the 70 percent threshold, it still carries a meaningful chance of refusal, especially if the conflicting mark is in a closely related class or if the marks share a dominant syllable. The safer approach is to modify the mark or narrow the description for any Medium Risk result where the conflicting mark is in the same or a related class. A second mistake is failing to consider the commercial impression of the marks as a whole. The similarity score measures phonetic, visual, and conceptual elements, but the examining attorney also evaluates the overall commercial impression. If your mark is "SunRise" and the conflicting mark is "SunRise Solutions," the score may be high, but the addition of "Solutions" may not be enough to avoid confusion if the goods are identical. In that case, modification is necessary.

How to Set Up Recurring Monitoring for a Multi-Brand Portfolio

For a multi-brand portfolio, you can set up recurring monitoring on AI Trademark Review by creating individual watch entries for each mark and class combination, with alerts consolidated to a single dashboard or email address. This method lets you track new USPTO applications and WIPO filings across all your brands without logging into separate accounts or running manual searches each week. The platform supports an unlimited number of watch entries under one account, so a portfolio of 50 marks in 10 classes each requires 500 separate entries — one per mark-class pair.

The workflow begins in the Monitoring section of your dashboard. Click "Add Watch," enter the exact mark string, and select the International Class numbers that apply. For a clothing brand, you would select Class 25. For a software brand, you would select Class 9 and Class 42. Repeat this process for every brand in your portfolio. Each watch entry runs independently, scanning the USPTO database for new applications filed within the last 7 or 30 days, depending on the alert frequency you choose. The default is weekly, which catches most new filings before they publish for opposition.

You can set the alert frequency per watch entry. Weekly alerts are sufficient for most portfolios because the USPTO publishes new applications every Tuesday and Thursday. Monthly alerts risk missing the 30-day opposition window for a conflicting mark. For high-value brands, set the frequency to weekly and enable push notifications to your mobile device. The platform sends a single digest email listing all new conflicts found across your portfolio, with similarity scores and risk ratings for each match. You can click directly from the email to view the full report.

For WIPO monitoring, navigate to the Monitoring section and select "WIPO" instead of USPTO. Enter your registered mark number — not the mark string — to receive notifications of new UDRP filings and international registrations that cite your mark. This is critical for brands with global distribution because a WIPO registration can block your U.S. expansion if the foreign mark has priority. You should set up WIPO watches for every mark that has an existing registration, not just pending applications.

One common mistake is creating a single watch entry for a mark without specifying the class. A watch entry with no class selected returns results across all 45 classes, which generates excessive false positives. For example, a watch on "Eclipse" without a class filter would alert you to every new application containing that word, from dental tools to software to energy drinks. Always pair each mark with its specific class numbers. Another mistake is forgetting to update watch entries when you add a new class to an existing registration. If you expand a clothing brand into Class 35 for retail services, you must create a new watch entry for that class.

To manage a large portfolio efficiently, use the portfolio dashboard view. This screen lists all your watch entries in a single table, showing the mark, class, last scan date, and number of new conflicts found. You can sort by last scan date to see which entries have not been reviewed. Export the full portfolio report as a CSV file each month for your records or to share with outside counsel. The CSV includes the mark, class, alert frequency, and a link to each conflict report.

Set aside 15 minutes each Monday morning to review the weekly digest. Open each conflict report, check the similarity score, and decide whether to file an opposition or a notice of intent to oppose. For scores above 70 percent in the same class, contact your attorney immediately. For scores below 30 percent, mark the entry as reviewed and move on.

What to do next

Your USPTO trademark search is the non-negotiable first step before filing. Now, use AI Trademark Review’s clearance tool to scan both USPTO and WIPO databases in a single query, then act on the risk scores to avoid office actions.

Also worth reading: Step-by-Step Guide Navigating the USPTO's TEAS System for Trademark Registration in 2024 · Navigating the New USPTO Trademark Search Tool A Step-by-Step Guide for 2024 · Trademark Registration Timeline Why Business Names Take 8-14 Months for USPTO Approval in 2024 · A Step-by-Step Guide to Using USPTO's TESS for First-Time Trademark Searches

Sources: uspto, usa, gov, europa