diff --git a/.gitbook/assets/uber_2fa.png b/.gitbook/assets/uber_2fa.png
new file mode 100644
index 0000000..3435e75
Binary files /dev/null and b/.gitbook/assets/uber_2fa.png differ
diff --git a/api-reference/commands/assertnodefectswithai.md b/api-reference/commands/assertnodefectswithai.md
new file mode 100644
index 0000000..80751b3
--- /dev/null
+++ b/api-reference/commands/assertnodefectswithai.md
@@ -0,0 +1,35 @@
+# assertNoDefectsWithAI
+
+{% hint style="warning" %}
+
+This is an **experimental** feature powered by LLM technology.
+
+{% endhint %}
+
+```yaml
+- assertNoDefectsWithAI
+```
+
+Takes a screenshot, uploads it to an LLM with a pre-made prompt, and asks the
+model if it sees any obvious defects in the provided screenshot. Common defects
+include text and UI elements being cut off, overlapping, or not being centered
+within their containers.
+
+### Output
+
+Output is generated in HTML and JSON formats in the folder for the individual
+test run:
+
+```
+~/.maestro
+└── tests
+    ├── 2024-08-20_213616
+    │   ├── ai-(My first flow).json
+    │   ├── ai-(My second flow).json
+    │   ├── ai-report-(My first flow).html
+    │   ├── ai-report-(My second flow).html
+```
+
+{% content-ref url="ai-configuration.md" %}
+[ai-configuration.md](ai-configuration.md)
+{% endcontent-ref %}
diff --git a/api-reference/commands/assertwithai.md b/api-reference/commands/assertwithai.md
new file mode 100644
index 0000000..f213af0
--- /dev/null
+++ b/api-reference/commands/assertwithai.md
@@ -0,0 +1,48 @@
+# assertWithAI
+
+{% hint style="warning" %}
+
+This is an **experimental** feature powered by LLM technology.
+
+{% endhint %}
+
+### When to use?
+
+{% hint style="warning" %}
+
+As all things generative AI, `assertWithAI` can be very helpful, but don't trust
+it blindly.
+
+{% endhint %}
+
+`assertWithAI` is useful when it's hard (or even impossible) to write the
+assertion using default assertion commands.
+
+Asserting the presence of a two-factor authentication prompt is a good example.
+
+<figure><img src="../.gitbook/assets/uber_2fa.png" alt=""><figcaption></figcaption></figure>
+
+```yaml
+- assertWithAI:
+    assertion: A two-factor authentication prompt, with space for 6 digits, is visible.
+```
+
+
+### Output
+
+Output is generated in HTML and JSON formats in the folder for the individual
+test run:
+
+```
+~/.maestro
+└── tests
+    ├── 2024-08-20_213616
+    │   ├── ai-(My first flow).json
+    │   ├── ai-(My second flow).json
+    │   ├── ai-report-(My first flow).html
+    │   ├── ai-report-(My second flow).html
+```
+
+{% content-ref url="ai-configuration.md" %}
+[ai-configuration.md](ai-configuration.md)
+{% endcontent-ref %}
diff --git a/api-reference/configuration/ai-configuration.md b/api-reference/configuration/ai-configuration.md
new file mode 100644
index 0000000..7f1c842
--- /dev/null
+++ b/api-reference/configuration/ai-configuration.md
@@ -0,0 +1,38 @@
+# AI features configuration
+
+Some commands, such as `assertWithAI` and `assertNoDiffWithAI`, use generative
+AI models, which are not built directly in Maestro CLI. Therefore, to use such
+commands, additional configuration is required.
+
+### Model
+
+The default model is the latest GPT-4o.
+
+You can configure the model to use with the `MAESTRO_CLI_AI_MODEL` env var, for example:
+
+```console
+export MAESTRO_CLI_AI_MODEL=claude-3-5-sonnet-20240620
+```
+
+Currently supported:
+– GPT family of models from OpenAI
+- Claude family of models from Anthropic
+
+Support for more models and providers is tracked [in this issue](https://github.com/mobile-dev-inc/maestro/issues/1957).
+
+### API key
+
+To use this command, an API key for the LLM service is required. To set it, export the
+`MAESTRO_CLI_AI_KEY` env var.
+
+For example, to set the key for OpenAI:
+
+```console
+export MAESTRO_CLI_AI_KEY=sk-4NXxdLXY4H9DZW0Vpf4lT3HuBaFJoz1zoL21eLoLRKlyXd69
+```
+
+or for Anthropic:
+
+```console
+export MAESTRO_CLI_AI_KEY=sk-ant-api03-U9vWi8GDrxRAvA2RL2RMCImYCQr8BFCbNOq2woeRXLNz2Iy4PbY1X2137leSm92mitI7F9IwxKIrXtXgTIzj7A-2AvgbwAA
+```
diff --git a/api-reference/configuration/flow-configuration.md b/api-reference/configuration/flow-configuration.md
index 71ae6ec..7b1d8d4 100644
--- a/api-reference/configuration/flow-configuration.md
+++ b/api-reference/configuration/flow-configuration.md
@@ -9,8 +9,7 @@ The following properties can be configured on a given Flow:
 * `onFlowStart`: This is a hook that takes a list of Maestro commands as an argument. These commands will be executed before the initiation of each flow. Typically, this hook is used to run various setup scripts.
 * `onFlowComplete`: This hook accepts a list of Maestro commands that are executed upon the completion of each flow. It's important to note that these commands will run regardless of whether a particular flow has ended successfully or has encountered a failure. Typically, this hook is used to run various teardown / cleanup scripts.
 
-```yaml
-# flow.yaml
+```yaml title="flow.yaml"
 appId: my.app
 name: Custom Flow name
 tags: