Menu

Methodology & datasets

 

A diagram showing the sources of our reports, the two workflows for assessing reports and imagery, and the IWF services they support.

 

IWF datasets:

This year, our Annual Data & Insights Report organises our datasets into clearly defined and distinct sections, each reflecting a specific way in which data informs and supports our operational work. These datasets consist of reports, URLs, and hashes, each serving a specific purpose within our response to online child sexual abuse imagery.

Reports

Reports account for a significant proportion of our work and are the primary mechanism through which content is brought to our attention. Most reports relate to URLs suspected of hosting child sexual abuse material. Reports may be submitted directly from external sources or generated internally to record our proactive work, when our analysts actively conduct searches online to identify and remove child sexual abuse. Since the launch of our child reporting services, Report Remove and Meri Trustline, we receive and assess reports submitted directly by children and young people through this same system.

When processing reports generated through our proactive activity, we record the URLs identified and then assess the content to determine whether any child sexual abuse material is present. This may include content directly displayed on a webpage, material accessed through links, referrals, advertisements, or paedophile manuals. We assess whether the material meets the threshold for child sexual abuse material. Throughout this report, references to “criminal” or "child sexual abuse” refer to the same definition.

URLs

URLs (or webpages) provide critical intelligence about where child sexual abuse imagery is being hosted online. Analysis of URL data allows us to identify hosting patterns, the types of websites involved, and those websites that are repeatedly found to host criminal content. In addition, this intelligence is valuable for identifying sites that are explicitly generating revenue from hosting this type of content.

Images and videos (imagery)

When we refer to direct content visible on a URL, we mean images, videos, or in some cases both, which we refer to collectively as imagery.

This imagery is uploaded into our internal system, Intelligrade, where assessors review each image and video. For images, assessors record the highest category of sexual activity and the age and sex of the youngest child involved. They then record the age and sex of any additional children appearing in the same image. Each image or video receives only one category assessment, regardless of any additional sexual activity present.

For individual items of imagery, we are able to record further details, referred to as metadata, such as the age and sex of each child and the type of sexual activity shown. Once an image has been fully assessed and the required information entered, a hash is automatically created. Identical images or videos share the same digital fingerprint, meaning that once a file has been hashed, duplicates do not need to be hashed again.

Hashes

A hash is a digital fingerprint of a file, such as an image or video. A hash is created by running a file through a special mathematical process that turns the file into a short string of letters and numbers. This string is unique to that file.

Even a tiny change to the file, like altering one pixel in an image, will produce a completely different hash. Because of this, hashes are used to quickly identify and match known criminal content online without needing to view the actual image or video itself.

Our hash dataset contains detailed information about individual images and videos, enabling analysis at a granular level. Hash data includes attributes such as the number of children depicted, estimated age and sex, the type of activity shown, and trend insights, including the identification of self-generated and AI-generated content. This approach allows for more precise trend analysis than URL-level data, where a single webpage may contain hundreds or thousands of images.

Within the hash dataset, images and videos are treated differently due to their complexity and impact on welfare. For images, we record the full range of available data. For videos, which present a higher welfare risk to analysts, we record severity alongside trend indicators relating to self-generated and AI-generated content.

IWF assessment and classification

Our work brings together action on publicly reported content, proactive detection of material hosted online, and the hashing and grading of imagery. We also process reports submitted through our child‑reporting tools, ensuring that children and young people can request the removal of their own intimate images. In addition, we grade images as part of our partnership with the UK Home Office’s Child Abuse Image Database (CAID), a secure national repository containing images and videos of child sexual abuse material collected by UK police forces and the National Crime Agency. This database plays a critical role in supporting investigations and securing convictions against offenders who create, access or distribute this material.

In 2025, we assessed more than 600,000 images and videos from CAID to determine whether they met the threshold for criminal classification. A significant proportion of these assessments resulted in non-criminal gradings. In many cases, the imagery did not meet the criminal threshold under UK law, or it was not possible to determine with complete confidence that a child was depicted.

We assess child sexual abuse material according to the levels detailed in the Sentencing Council's Sexual Offences Definitive Guideline The Indecent Photographs of Children section (Page 34) outlines the different categories of child sexual abuse material.


Category A: Images involving penetrative sexual activity; images involving sexual activity with an animal; or sadism.
Category B: Images involving non-penetrative sexual activity.
Category C: Other indecent images not falling within Categories A or B.

Prohibited images are assessed under a separate legal framework (Section 62 of the Coroners and Justice Act 2009) to indecent images that fall under the Protection of Children Act 1978 and Section 160 of the Criminal Justice Act 1988. Prohibited images are non-photographic images (including computer generated images (CGI), cartoons, manga images and drawings)

Dataset development and welfare considerations

We continue to seek opportunities to expand our datasets and share insights with industry and partners. However, this must be carefully balanced against the welfare of our analysts and assessors. Our assessment work on both CAID-sourced and proactively found imagery prompted us to change how we respond to imagery that depicts child exploitation, but doesn't explicitly depict child sexual abuse according to UK laws. As a result, we introduced an ‘exploitative’ category to better reflect and respond to this type of content.

Exploitative category:

Exploitative content refers to any material, particularly images, that depicts or implies the sexualisation or victimisation of a minor, even where the content itself may fall short of the legal threshold for criminality. 

This includes, but is not limited to:

Borderline criminal sexualised depictions of a child
Content that does not meet the threshold for illegality under UK law,  but is still sexualised in its nature or intent

Images linked to known exploitation
Lawful images of a child that, in the appropriate context, are linked to imagery of known or suspected sexual exploitation of the same child. These links can be determined via victim identification, distribution patterns and metadata.

Images believed to depict a child but without confirmed identification
Content where there is high confidence that a child is depicted, but where it is not possible to confirm this with complete certainty without independent confirmation.

We classify these areas of content as exploitative because it contributes to, or risks contributing to, the sexual exploitation or continued harm of a child, even where a single image cannot be graded as criminal.

In October 2025, we began the internal process of grading exploitative images which included reassessing a number of images that had previously been assessed as lawful.

In just over two months, our team assessed more than 72,000 images as belonging to the exploitative category.

Exploitative breakdown

  • Borderline indecent
  • Known or confirmed victims
  • Age in question
  • Borderline indecent & age in question

While this category was initially developed for internal use, we intend to expand its application and share this intelligence with industry and partners. Deployed through our membership services, the industry application of blocking Exploitative imagery could strengthen protections for victims, and potentially support earlier intervention on the uploading of criminal content. By sharing this insight more broadly, we can:

  • Support online platforms to improve detection systems, strengthen safety-by-design measures, and respond more quickly to emerging risks.

  • Inform government and policymakers to help shape proportionate regulation, guidance and preventative strategies.

  • Enable civil society and child protection organisations to better understand emerging harm patterns and tailor prevention and support services.

Our Hotline team

Internet Content Analysts

The Internet Content Analysts, often referred to as Analysts, are a team of 17 people. They are responsible for proactively searching for images and videos of child sexual abuse online, responding to public reports, and monitoring new trends. It’s their job to ensure criminal content depicting the sexual abuse of children is removed from the internet.  

Image Classification Assessors

The Image Classification Assessors, often referred to as Assessors, are a taskforce of 14 whose role is to assess images and videos, adding extra metadata to each image such as the age of the child depicted and the type of sexual activity that is occurring, in addition to other information. Once the data is added, a hash or “digital fingerprint” is created. These hashes are then used to prevent the upload, download and further dissemination of this image by our industry partners. 

Quality Assurance Assessors and Officers

The role of our Quality Assurance team is to support the Hotline. The team of five are intentionally managed by a different Director to the rest of the Hotline and they ensure the work of the Hotline is held to the highest standards. They check for accuracy and consistency of assessments and track trends to ensure the IWF remains a trusted and world-leading organisation.