How file formats and Add-Ins affect Content Risk Reviews

Modified on Thu, 23 Apr at 1:44 AM

A guide to how file formats and Add-Ins affect compliance scanning.

Key principle

A well-configured rule cannot detect something the extractor did not read correctly. Format changes are the most common cause of rules not firing as expected.

Our platform uses AI to check every marketing asset for legal, regulatory, and brand compliance. Before any rule can fire, the platform must first read and extract the content from your file — a step called content extraction. The way your file is structured directly affects how well this extraction works, and therefore how reliably your compliance rules fire.

How we read your assets

Each file type is processed differently when uploaded to the platform or sent to IntelligenceBank via the Add-In. Understanding which method applies to your assets helps you anticipate where rule detection may be stronger or weaker.

PDF

PDF — text-based

Text is extracted directly from the PDF layer. This gives the highest rule detection accuracy. Ensure your PDF is exported with a proper text layer, not flattened as an image.

Image

Image / PNG / JPG

Text is read using OCR (optical character recognition). Accuracy depends on image resolution, font legibility, and contrast between text and background.

HTML

Web page / HTML

Text is extracted from the page structure. Multi-column or complex layouts can affect the order in which content is read and matched against rules.

PPTX

PowerPoint / PPTX

Slide text is extracted from text boxes. Images or diagrams embedded inside slides are processed via OCR and may have lower detection accuracy.

DOCX

Word / DOCX

Text is extracted from the document structure. Columns, tables, and floating text boxes affect reading order and can impact rule detection.

The layout change problem

A dealer finance web page may be set up as a two-column layout, with compliance rules trained and validated against that format. When the same campaign content is produced as a social image (PNG) or a PDF poster, the content extraction process changes completely. Rules that fire reliably on the web page may not fire on the image.

Web page — 2 columns

Content is extracted column by column. Rules can detect rate disclosures, comparison rates, and end-date text correctly because the content is machine-readable.

Social image (PNG)

The same content must be read by OCR. If resolution is low or text overlaps a background image, OCR may miss the comparison rate disclosure — and the rule does not fire.

Web page — 3 columns

Adding a third column changes the extraction order. Rules trained on the two-column output may no longer match the new structure, and disclosures can appear in unexpected positions.

PDF poster

A text-layer PDF gives high accuracy. If the PDF was exported as a flattened image, it is treated as a PNG and OCR is applied, which reduces detection accuracy.

Tell us when your asset format changes

Your compliance rules are calibrated to your specific asset types and layouts. When those change, rules may need to be reviewed or updated. Please notify your Customer Success Manager whenever any of the following occur:

● You change the column layout of a webpage or document, such as moving from two columns to three.

● You add a new asset type to your workflow, such as social images when you previously only produced web pages.

● You change from a text-layer PDF to a flattened or image-based PDF, or vice versa.

● You change image dimensions, DPI, or resolution settings on image assets.

● You switch from one design template to another with different text placement.

● You add or remove disclaimer or disclosure sections from a standard layout.

Good practice

Whenever you create a new campaign template or introduce a new asset format, share one example with your CSM before going live. A quick format review takes far less time than investigating missed rules after a campaign has launched.

What to do before uploading a new asset

Identify the change

Note what is different — column count, file type, template design, or export settings.

Collect a sample

Gather at least 3 examples of the new asset type before reaching out to your CSM.

Notify your CSM

Share the samples and a brief description of what changed and which rules should apply.

Wait for confirmation

Do not use the new format in production until your CSM confirms rules have been reviewed.

Resume as normal

Once confirmed, upload freely — your rules now reflect the updated format.

Add-in Reviews

What to expect from off-platform Add-In reviews

Add-In reviews are a great way to use your Content Risk Review rules in an agile off-platform experience from where you create content. Rule results help authors refine their content before doing on-platform review and approval. There are some nuances to keep in mind when using the Add-Ins.

PowerPoint Add-In

PPT(X) reviewed in PowerPoint

Text and images are reviewed, and OCR is applied to text embedded in images. Rules are annotated on a per-slide basis, and individual risk words are not directly highlighted.

Word Add-In

DOC(X) reviewed in Word

Document body text is reviewed. Headers, footers, images, and text embedded in images are not reviewed. Rules are annotated in the document with risk word highlighting.

Figma

FIG files reviewed in Figma

Selected frames are converted to PDF for review.

Universal Connector / GenStudio Add-In

Content reviewed in Universal Connector

Users can review:

Uploaded files in the formats and sizes supported by the API
Selected text copy
Web pages automatically saved as PDFs for review

Connected Apps generally send files for review, though some may use other content types such as pages or content fragment data for text review.

Questions?

Contact your Customer Success Manager. This guide should be reviewed whenever your standard production asset mix changes.