Today's AI Tip
← All posts

I asked AI to remove duplicate contacts. It quietly dropped 4 emails.

· View on Instagram · View on Facebook

I asked AI to remove duplicate contacts. It quietly dropped 4 emails.
The prompt most of us type. We pasted a fictional 20-row list and typed "Remove the duplicate contacts from this list and give me the clean list." GPT-5.5 returned 14 rows and no explanation.

The prompt most of us type

Real output · GPT-5.5 (OpenAI API) · Oct 2026

```csv Name,Company,Email,Phone,City,Title John Smith,Brightline Design,john.smith@brightline.example,+44 113 496 0101,Leeds,Account Director Priya Shah,Harbor Foods,priya.shah@harborfoods.example,+44 117 496 0202,Bristol,Buyer …

We pasted a fictional 20-row list and typed "Remove the duplicate contacts from this list and give me the clean list." GPT-5.5 returned 14 rows and no explanation.

What disappeared without a word. In our test, 6 rows merged away and 4 email addresses vanished, plus Maria Garcia's other company and title. Nothing in the output told us.

What disappeared without a word

In our test, 6 rows merged away and 4 email addresses vanished, plus Maria Garcia's other company and title. Nothing in the output told us.

Source: Our test: GPT-5.5 (OpenAI API), Oct 2026

Ask for an audit trail. Anthropic's guide suggests making answers auditable by having the model cite support for each claim. Here, that means every merge shows the rows it came from.

Ask for an audit trail

Find the duplicate contacts in this list. For every merge, show both original rows in full, say which fields matched and which differed, and say which value you kept for each field that differed. Only merge rows that clearly describe the same person. If two rows might be the same person but you are not sure, list them under Check by hand and do not merge them. Then give the clean list and the row count before and after.

Tested in GPT-5.5 (OpenAI API), Oct 2026

Anthropic's guide suggests making answers auditable by having the model cite support for each claim. Here, that means every merge shows the rows it came from.

Source: Anthropic, Reduce hallucinations (Claude docs)

Showing the work didn't fix it. We got the same 6 merges, but now we could check them. Two Grace Okafor rows matched only on name and city, and were merged anyway.

Showing the work didn't fix it

Real output · GPT-5.5 (OpenAI API) · Oct 2026

### Merge 3: Maria Garcia duplicate … Because there is no date or freshness indicator, I kept the first full work record. … ### Merge 5: Grace Okafor duplicate … **Matched fields:** Name, City. The second row is sparse and has no conflicting company, phone, or title.

We got the same 6 merges, but now we could check them. Two Grace Okafor rows matched only on name and city, and were merged anyway.

Write the merge rule. A rule changed the decision. Names must match, and so must the email or phone. Conflicting values both stay, and every other possible match goes to Check by hand.

Write the merge rule

Find the duplicate contacts in this list. Merge two rows only if the names match (allowing nicknames) AND the email or the phone number matches too. For every merge, show both original rows in full. If two values conflict, keep both (add an Email 2 or Note column) instead of choosing one. List every other possible match under Check by hand and do not merge it. Then give the clean list and the row count before and after.

Tested in GPT-5.5 (OpenAI API), Oct 2026

A rule changed the decision. Names must match, and so must the email or phone. Conflicting values both stay, and every other possible match goes to Check by hand.

Fifteen rows, every email kept. In our test, the rule gave 15 rows with every email kept, and the Grace Okafor pair went to Check by hand. The two David Chens stayed separate.

Fifteen rows, every email kept

Real output · GPT-5.5 (OpenAI API) · Oct 2026

## Row count - Before: 20 rows - After: 15 rows … | 15 & 18 | Grace Okafor / Grace Okafor | Same name and city, but email does not match and one phone is blank. |

In our test, the rule gave 15 rows with every email kept, and the Grace Okafor pair went to Check by hand. The two David Chens stayed separate.

Where it still needs you. A shared phone isn't proof: Tom Reed and Anna Kowalski share an office number, and the model flagged it. Maria Garcia's row still lists Oakfield Partners, and only a person knows her current job.

Where it still needs you

A shared phone isn't proof: Tom Reed and Anna Kowalski share an office number, and the model flagged it. Maria Garcia's row still lists Oakfield Partners, and only a person knows her current job.

Try it: Run it on a 20-row sample first and check every merge.

Source: Our test: GPT-5.5 (OpenAI API), Oct 2026

Before you paste a real list. Remove names, account numbers and other personal details first, and check the tool's data policy or your employer's approved tools before uploading real documents.

Before you paste a real list

Remove names, account numbers and other personal details first, and check the tool's data policy or your employer's approved tools before uploading real documents.

Sources and assumptions

Assumptions:

The short version

Follow @todaysaitipSave this post on Instagram

More posts