> For the complete documentation index, see [llms.txt](https://docs.uptiq.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.uptiq.ai/document-ai/cookbooks/identity-verification.md).

# Verify an identity document

**Goal.** Extract the fields from a passport, driver's licence or other identity document, and use the result — together with its integrity signals — to support a KYC check.

### Before you start

* An API key, exported as `$DOCAI_API_KEY`.
* An identity document image or PDF.

{% hint style="danger" %}
Identity documents are personal data of the most sensitive kind. Everything here happens inside your regulated pipeline: log field values only where your retention policy allows, never write them to application logs by default, and delete source images on the schedule your policy sets. Nothing in this recipe should be pasted into a shared notebook or a support ticket.
{% endhint %}

### Step 1 — Find the right document type

Identity documents form their own category in the registry:

```bash
curl -s "https://api-docai-uat.uptiq.ai/listSupportedDocuments" \
  | jq -r '.documents[] | select(.category == "Identity Document") | "\(.type)\t\(.name)"'
```

```
AadhaarCard     Aadhaar Card
Passport        Passport
…
```

See [Document Types](/document-ai/guides-1/document-types.md) for the full categorised list.

### Step 2 — Check the accepted formats first

ID documents are usually photographed rather than scanned, and the accepted formats differ from financial documents — spreadsheets are accepted for a bank statement and meaningless here.

```bash
curl -s "https://api-docai-uat.uptiq.ai/document/supported-file-types" \
  -H "X-Api-Key: $DOCAI_API_KEY" | jq '.data.AadhaarCard'
```

```json
["pdf", "jpg", "jpeg", "png"]
```

Validating the extension before you submit saves a wasted call and a wasted credit.

### Step 3 — Extract

If you know which document you were given, name it:

```json
{ "documentType": "Passport", "content": "…" }
```

If the customer simply uploaded "ID", classify first — a passport and a driver's licence carry different fields, and extracting one as the other returns very little:

```bash
curl -X POST "https://api-docai-uat.uptiq.ai/classify/sync" \
  -H "X-Api-Key: $DOCAI_API_KEY" \
  -H "Content-Type: application/json" \
  --data-binary @classify.json
```

Classification runs in around 13 seconds, so this is affordable in an interactive flow in a way that extraction is not. See [Classify, then extract](/document-ai/cookbooks/classify-then-extract.md).

{% hint style="info" %}
`subtype` is worth reading on the result. The webhook contract shows `"type": "Driver's License"` with `"subType": "CA"` — the issuing jurisdiction, which usually determines which validation rules apply.
{% endhint %}

### Step 4 — Judge the capture quality

Photographed IDs fail differently from scanned documents — glare, angle, a thumb over the corner. `documentQuality` tells you whether a thin result is the image's fault:

```bash
jq '.extraction.result.documentQuality' record.json
```

```json
{ "level": "high", "sharpness": 833.29, "stage": "extraction", "reasons": [] }
```

Low `level`, or a populated `reasons` array, means ask for a better photograph rather than retrying the same one or escalating to a human. That single decision removes most of the avoidable review load in an onboarding flow.

### Step 5 — Check the integrity signals

For identity documents this is not optional — a forged ID is the threat the whole check exists for.

```bash
jq '.extraction.fraudResults.preflight.results[]
    | select(.status == "FAIL")
    | {ruleId, category, explanation: .details.explanation}' record.json
```

`visual_forensic` is the category that matters most here: it looks for tampering evidence such as inconsistent fonts, edited regions and missing print artifacts.

{% hint style="warning" %}
Two traps, both covered in [Reading fraud and control checks](/document-ai/cookbooks/fraud-and-control-checks.md):

**Fraud analysis finishes after extraction does.** An extraction at `Processed` may still have `fraudStatus: "processing"`. For KYC, fail closed — hold the case rather than clearing it on an incomplete signal.

**Gate on `failed`, not on `score`.** A category score can be low simply because rules were *skipped* for lack of evaluable evidence. "Could not assess" is a review decision, not a pass and not a rejection.
{% endhint %}

### Step 6 — Decide

| Signals                                                                 | Decision                                                              |
| ----------------------------------------------------------------------- | --------------------------------------------------------------------- |
| Fields complete, quality high, no failed rules, fraud analysis terminal | Pass to your matching step                                            |
| Quality low                                                             | Request a better image — do not escalate to a human yet               |
| Fields missing but quality high                                         | Human review; possibly the wrong document type                        |
| Any `visual_forensic` failure                                           | Hold for the fraud team. Do not auto-reject on a machine signal alone |
| Fraud analysis not terminal                                             | Wait, then decide. Never clear on a partial result                    |

Extraction supports the check; it does not make the decision. Whether the extracted name matches your applicant, whether the document has expired, and whether the jurisdiction is acceptable are all your system's calls.

### What is not here

The API extracts and reports integrity signals. It does **not** do biometric matching, liveness detection, or sanctions and PEP screening — those come from your KYC provider. Nor can fraud rules be configured through the API; that is [Fraud Detection](/document-ai/guides-1/fraud-detection.md) in the portal.

### Related pages

* [Reading fraud and control checks](/document-ai/cookbooks/fraud-and-control-checks.md) — the integrity signals in detail.
* [Classify, then extract](/document-ai/cookbooks/classify-then-extract.md) — when you do not know which ID you were given.
* [Document Types](/document-ai/guides-1/document-types.md) — the Identity Document category.
* [Fraud Detection](/document-ai/guides-1/fraud-detection.md) — configuring the rules behind the checks.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.uptiq.ai/document-ai/cookbooks/identity-verification.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.