# AI Duplicate Matching – Definition, FAQs & How Dedupely Helps > AI duplicate matching explained: what it is, how it finds duplicate records that matching rules miss, and how to review its matches before merging. - URL: https://dedupe.ly/glossary/ai-duplicate-matching - Section: glossary - Site index: https://dedupe.ly/llms.txt - Any page: append `.md` to its URL for this plain-text form --- [Glossary](https://dedupe.ly/glossary) AI Duplicate Matching ## AI Duplicate Matching AI duplicate matching is a way of finding duplicate records in which artificial intelligence is used to decide which records describe the same person, company or deal. It runs many sets of comparisons at once, so it can catch duplicates that only show up when several small signals line up: a similar name, a phone number in another format and a shared company domain, for example. Rule-based matching finds the duplicates someone thought to look for. Each rule names the fields to compare and how strictly to compare them, so the duplicates that fall between rules stay in the CRM. AI duplicate matching is built for that remainder. It covers many field combinations no one wrote a rule for, and it needs little or no setup. Its matches still go to a person, or through merge rules, before anything is merged. See how Dedupely Intelligence finds the duplicates your searches miss [Try for free](https://v2.app.dedupe.ly/signup) [Talk to sales](https://meetings.dedupe.ly/meetings/gaby-salinas/talk-to-sales) Overview AI duplicate matching uses artificial intelligence to decide which CRM records are duplicates. It compares many fields and field combinations at once, so it catches duplicates that hand-written rules miss. It works best as a second pass after rule-based matching, with a person reviewing matches before they're merged. On this page - What Is AI Duplicate Matching? - How AI Duplicate Matching Works - Why AI Duplicate Matching Matters for CRM Teams - Getting Started With AI Duplicate Matching - Key Takeaways: AI Duplicate Matching - Frequently Asked Questions About AI Duplicate Matching - Related Terms and Concepts See how Dedupely Intelligence finds the duplicates your searches miss [Try for free](https://v2.app.dedupe.ly/signup) [Talk to sales](https://meetings.dedupe.ly/meetings/gaby-salinas/talk-to-sales) ### What Is AI Duplicate Matching? AI duplicate matching is a record-matching method in which AI shapes the decision about which records belong together. A typical example: "J. Smith, Acme, (555) 010-2000" and "John Smith, Acme Inc, 555-010-2000, john@acme.com." No single field matches exactly, so a rule on email or full name misses the pair. Taken together, the name, company and phone number point to the same person, and AI duplicate matching weighs all three at once. The duplicates it's built to catch are the ones that differ in several small ways: abbreviated names, legal suffixes on company names, phone numbers with and without country codes, values typed into the wrong field, and email addresses on different domains for the same company. It applies to contacts, companies, leads and other records, in any CRM. There are two ways to build it. In the first, AI is used ahead of time to design the matching logic: which fields to compare against each other, and how. That logic then runs as fixed code, so the same records always produce the same matches, and no record is sent to an AI model. In the second, records are sent to an AI model, often an outside large language model, which compares them while it runs. CRM data leaves the system, and results can change between runs. The difference matters for privacy and for how predictable the results are. Resources: - [Match Options](https://dedupe.ly/knowledge-base/match-options) (KB) - [How to correctly identify all duplicates in your CRM](https://dedupe.ly/blog/merge-with-confidence) ### How AI Duplicate Matching Works AI duplicate matching follows the same two stages as any deduplication, matching and then merging, but it changes how the matching stage is set up. - **It runs many comparisons at once.** Instead of one rule per pattern, it compares many fields against each other, such as a name against an email address, or a website against an email domain, so a match doesn't depend on one field a person picked. - **It counts several kinds of similarity.** Exact values, typos, names that sound alike, phone numbers in different formats and shared root domains can all count, and several small similarities together can point to a strong match. - **It groups the matches into duplicate sets.** Records that point to the same person, company or deal end up in the same set, ready for a person to review before anything is merged. - **Merging still follows the usual rules.** Merge rules decide which record is kept and which value wins in each field, the same way they do for rule-based matches. Where the AI does its work depends on the tool. When AI designs the comparisons ahead of time, these steps run as fixed code, the same records always give the same matches, and no record is sent to an AI model. When records are sent to an AI model, the model makes those comparisons while it runs. Either way, there are few or no match settings to configure, and a person still reviews the matches before anything is merged. Resources: - [The limits of native deduplication in CRMs](https://dedupe.ly/blog/the-limits-of-native-deduplication-in-crms) ### Why AI Duplicate Matching Matters for CRM Teams The duplicates left after a rule-based cleanup are often the hardest ones: records that differ in small ways across several fields at once. Finding them by hand means writing, testing and combining many searches, one for each pattern, and many teams stop once the obvious matches are gone. The duplicates that remain keep inflating contact counts, splitting activity history across records and putting the same customer in front of two reps. AI duplicate matching can close much of that gap with little setup. A team that would need a dozen carefully tuned searches to cover every pattern can review one list of matches instead, and spend its time deciding what to merge rather than building rules. That matters most in large or older CRMs, where years of imports, form fills and syncs have left duplicates in patterns nobody remembers creating, and where a rule for each one would take weeks to write and test. It also raises questions worth answering before turning it on: where the CRM data is processed, whether the same records always give the same result, and whether anything is merged without a person approving it. The answers depend on which kind of AI duplicate matching a tool uses, so they're worth checking before connecting a CRM. Resources: - [Why your CRM database is turning into a mess (and how to fix it)](https://dedupe.ly/blog/why-your-crm-database-is-turning-into-a-mess) ### Getting Started With AI Duplicate Matching Use AI duplicate matching as a second pass. Run your rule-based searches and merge the clear duplicates first, so the AI matches are the hard-to-spot ones that are left, and the list stays short enough to review. Before merging its matches, check where the data is processed and whether the same records always give the same result. Set the rules that decide which record and values are kept, and review each set before it's merged. Matches that aren't real duplicates should be rejected, so they don't come back in the next scan. Keep your own rules for the patterns you already know. They stay easy to explain and can be automated, while AI duplicate matching covers the combinations nobody wrote a rule for. Start with a small batch of its matches, such as one object at a time. Look at why each set was grouped, merge the ones you're sure about and reject the rest, before working through the full list. In Dedupely, AI duplicate matching is called Dedupely Intelligence. It runs as fixed matching logic on Dedupely's own infrastructure, so CRM records are never sent to outside AI models, and it appears at the top of your Duplicate Searches with nothing to set up. Resources: - [Dedupely Intelligence](https://dedupe.ly/knowledge-base/dedupely-intelligence) (KB) - [Duplicate Searches](https://dedupe.ly/knowledge-base/search-pads) (KB) ### Key Takeaways: AI Duplicate Matching AI duplicate matching finds the duplicates that rule-based matching misses by comparing many fields and field combinations at once, with little or no setup. It works best as a second pass after rule-based searches, and its matches still go through review and merge rules before anything changes in the CRM. Before using it, check where the data is processed and whether results are repeatable. Dedupely Intelligence is Dedupely's AI duplicate matching: it runs as fixed logic on Dedupely's own infrastructure, covers Contacts, Companies and Leads, and uses your Merge Rules and Safety Settings. Duplicate Searches cover the other objects and the automation, through Auto Match and Auto Merge. Resources: - [Dedupely product overview](https://dedupe.ly/product) ### Frequently Asked Questions About AI Duplicate Matching #### Does AI duplicate matching send CRM data to an outside AI model? It depends on the tool. Some AI duplicate matching sends records to a model while it runs, which can include an outside large language model, so it's worth asking a vendor where the data is processed, whether it's stored, whether it's used for training, and whether results can change from one run to the next. Fixed matching logic avoids those questions, because nothing leaves the matching system. Dedupely Intelligence works that way. AI was used only in Dedupely's development environment to design the matching logic. When it runs on your CRM, it's that fixed logic running on Dedupely's own infrastructure, and CRM records are never sent to, processed by or used to train third-party AI models or external LLMs. Inside the app, that logic runs as standard deterministic code. #### AI duplicate matching vs. rule-based matching: what changes? Rule-based matching compares the fields you choose with the method you choose, such as exact, similar or phonetic matching on name and email. It's precise and easy to explain, but it only finds the duplicates the rules describe, and every new pattern needs a new rule. AI duplicate matching compares many fields and field pairings at once and weighs several weak signals together, so it finds duplicates no single rule would catch, with little or no setup. Many teams use both: rules for the patterns they know and want automated, AI duplicate matching for the rest. In Dedupely, Duplicate Searches with Match Options handle rule-based matching and automation on the objects you sync, and Dedupely Intelligence runs separately on Contacts, Companies and Leads. It doesn't change or limit the Duplicate Searches you've set up. #### Why does AI duplicate matching find duplicates my own searches missed? Because your searches only look for the patterns you defined. An exact search on name and phone number won't pair a record that has a shortened name, a different phone format and a work email with one that has the full name and a personal email, even though they're the same person. AI duplicate matching compares all of those fields together, so it catches combinations nobody wrote a rule for. Because Dedupely Intelligence evaluates several field combinations at the same time, it can find complex, overlapping matches that single-rule searches miss, including duplicates that only appear when three or four small differences line up. It also runs separately from your own searches, so those matches show up in their own list instead of being added to yours. If you see this, you can merge those sets from Dedupely Intelligence, or add a Duplicate Search with the right Match Options so the same pattern is covered by your own rules, including Auto Match and Auto Merge. #### Can AI duplicate matching merge duplicates automatically? Some tools offer it, but it's safest to review AI matches before merging, because merges are hard or impossible to reverse in most systems. In Dedupely, merges can't be undone, so Auto Match and Auto Merge are turned off in Dedupely Intelligence on purpose, and there's no option to select every match across the database at once. You review the duplicate sets and merge them yourself, and each merge follows your Merge Rules and Safety Settings, so the right record and the right values are kept. For duplicates you want handled automatically, set up a Duplicate Search with the right Match Options and turn on Auto Match and Auto Merge there: Auto Match notifies you when new matches reach a count you set, and Auto Merge merges them on your schedule. #### Do you still need your own matching rules if you use AI duplicate matching? Yes, in most cases. AI duplicate matching finds the hard-to-spot duplicates, but your own rules decide what happens every day: which objects are covered, which duplicates are merged automatically and which you review yourself. In Dedupely, Dedupely Intelligence currently covers Contacts, Companies and Leads, so duplicates in objects like deals, tickets or custom objects still need Duplicate Searches. Automation also lives in your own searches, since Auto Match and Auto Merge are turned off in Dedupely Intelligence. Your Merge Rules matter more, not less, because Dedupely Intelligence uses them to decide which record is kept and which values win when you merge its matches. Without them, merges follow your CRM's native merge logic. The two work best together: AI duplicate matching shows you patterns you hadn't spotted, and your rules turn the ones that keep coming back into searches that run on their own. ### Related Terms and Concepts #### [CRM deduplication](https://dedupe.ly/glossary/crm-deduplication) CRM deduplication is the full process AI duplicate matching is part of: finding duplicate records, merging them, and catching new ones as they arrive. AI duplicate matching changes how the finding is done; the merging works the same way. #### [Matching duplicate records](https://dedupe.ly/glossary/matching-duplicate-records) Matching duplicate records is the stage of deduplication where records are compared to decide which ones describe the same thing. AI duplicate matching is one way to do it, alongside rule-based methods like exact and similar matching. #### [Fuzzy matching](https://dedupe.ly/glossary/fuzzy-matching) Fuzzy matching catches names that sound alike, one of the signals AI duplicate matching weighs. It often compares one field with one method; AI duplicate matching combines it with many other signals. #### [Multi-field matching](https://dedupe.ly/glossary/multi-field-matching) Multi-field matching combines several fields in one rule, such as name + company + phone. AI duplicate matching takes the same idea further by comparing many field combinations without each one being written as a rule. #### [False matches](https://dedupe.ly/glossary/false-matches) False matches are records paired as duplicates that aren't. They're the main reason AI matches are reviewed before merging, and why rejected matches should stay out of future results. #### [Deduplicating data with AI](https://dedupe.ly/glossary/deduplicating-data-with-ai) Deduplicating data with AI covers every way AI is used to find duplicates, from asking a general AI assistant to check an exported file to matching built into a deduplication tool. AI duplicate matching is the second kind: it works on the records in the CRM, and its matches can be reviewed and merged there. Skip to content