Automatically Skip Telegram Posts with Image Keywords Using OCR AI
Use OCR Image AI to detect a keyword inside an image, add a unique marker to the message content, and use Blacklist to skip the post automatically.
Download Mobile App or use Web
✅ iOS → App Store ✅ Android → Google Play ✅ Web → web.autoforwardtelegram.com
What does this feature do?
This workflow is useful when you want to ignore posts that contain a specific word or phrase in an image.
For example, you can detect the English word PROMOTION and skip any post whose image contains that word:
OCR Image AI reads the image.
The custom prompt returns a unique marker.
APPEND TO CAPTION adds the marker to the forwarded message content.
Blacklist detects the marker and prevents the post from being forwarded.
Posts without the keyword continue through the forwarding workflow.
Before you start
Create or open the forwarding task that contains the source and destination chats.
Prepare a clear test image containing the keyword
PROMOTION.Use a unique marker that is unlikely to appear in normal captions.
Allow a little extra time for OCR processing. The app notes that using AI may cause slight delays in message forwarding.
1. Enable OCR Image AI
Open the OCR settings for your forwarding task.
Turn on OCR Image AI Enable.
Select Custom Prompt.
Keep your preferred model selected in Model.
2. Add the keyword prompt
In Prompt Request, enter the following prompt:
The prompt returns OCR_SKIP_PROMOTION only when the keyword is found. Using a unique marker makes the filter more precise than adding the common word PROMOTION directly to Blacklist.
3. Append the OCR result to the message
Under Text Output Position, select APPEND TO CAPTION.
This setting adds the text extracted from the image to the end of the forwarded message content. In this example, a matching image adds OCR_SKIP_PROMOTION to the content that Blacklist can inspect.
Click Save & Apply.
4. Add the marker to Blacklist
Open Blacklist in the same forwarding task and add this exact value:
Save the Blacklist settings.
Do not add OCR_PASS or other generic words to Blacklist. A unique marker reduces the risk of skipping unrelated posts.
5. Test the workflow
Test an image that contains the keyword
Send a new post to the source chat. Use a neutral caption and an image that clearly contains PROMOTION.
Expected result:
OCR detects
PROMOTION.The custom prompt returns
OCR_SKIP_PROMOTION.APPEND TO CAPTION adds the marker to the message content.
Blacklist matches the marker.
The post is skipped and does not arrive in the destination chat.
Test an image without the keyword
Send another new post with an image that does not contain PROMOTION.
Expected result:
The marker is not returned.
Blacklist does not match.
The post is forwarded normally.
Use new test posts for each check so that you can clearly identify the result of the current configuration.
After setup
The task is ready to skip image posts that contain PROMOTION. The OCR setting and the Blacklist entry work together:
Contains PROMOTION
OCR_SKIP_PROMOTION
Match
Skipped
Does not contain PROMOTION
Empty response
No match
Forwarded
Manage the configuration
Change the keyword
To detect a different word:
Update the keyword and marker in Prompt Request.
Select APPEND TO CAPTION.
Click Save & Apply.
Replace the old marker in Blacklist with the new marker.
Test the new keyword with a new image.
For example, replace PROMOTION with DISCOUNT and OCR_SKIP_PROMOTION with OCR_SKIP_DISCOUNT.
Disable OCR filtering
Turn off OCR Image AI Enable when you no longer want the task to analyze images. If the Blacklist entry remains, remove it separately if you no longer need that filter.
Remove the filter but keep OCR
Remove OCR_SKIP_PROMOTION from Blacklist. OCR can remain enabled, but matching posts will no longer be skipped by this Blacklist rule.
Common statuses
OCR enabled
Image analysis is active for the task.
Confirm the custom prompt and output position.
Custom Prompt selected
OCR follows the instructions in Prompt Request.
Check the keyword and marker spelling.
APPEND TO CAPTION selected
OCR output is added to the forwarded message content.
Keep this selected when Blacklist must inspect the OCR result.
Blacklist marker present
Matching content is filtered.
Use the exact marker returned by the prompt.
Troubleshooting
A post containing the keyword is still forwarded
Check the following:
OCR Image AI Enable is turned on.
Custom Prompt is selected.
The keyword and marker are spelled exactly the same in the prompt and Blacklist.
APPEND TO CAPTION is selected.
You clicked Save & Apply after changing the OCR settings.
The test uses a new post and a clear image.
Allow a short delay for OCR processing before checking the destination chat.
Posts without the keyword contain unexpected OCR text
The model may return text even when the keyword is not present. Make the prompt stricter and require an empty response when there is no match. If your model cannot reliably return an empty response, use a non-blacklisted value such as OCR_PASS; note that this value may be appended to the forwarded caption.
OCR does not detect the keyword
Use an image with larger, sharper, high-contrast text. Avoid heavily blurred, rotated, or partially hidden words. Test the exact spelling used in the prompt.
Unrelated posts are skipped
Use a more distinctive marker, such as OCR_SKIP_PROMOTION, instead of adding a common word directly to Blacklist. Also check that normal captions do not already contain the marker.
Frequently asked questions
Can I use a phrase instead of a single word?
Yes. Replace PROMOTION in the prompt with the phrase you want to detect and update the marker in Blacklist.
Do I need to add the keyword itself to Blacklist?
No. The recommended setup adds a unique OCR marker, such as OCR_SKIP_PROMOTION, to Blacklist. This helps prevent false matches in normal captions.
Will OCR slow down forwarding?
AI image processing may cause slight delays in message forwarding.
What happens when the OCR setting is disabled?
The task no longer uses OCR Image AI for this filter. The Blacklist rule is managed separately and can be removed if it is no longer needed.
Last updated