Author and Image Smart Categories

Smart Categories can analyze more than just post text. With image and video categorization and author data, you can build classification rules that reflect what’s actually shown in visual content and who is driving a conversation — not just the words in a caption.

​

Image and video Smart Categories

​

The basics

What is image categorization in Smart Categories? Image categorization extends the same AI intelligence that Smart Categories apply to text — to visual content. When a mention includes an image, the platform analyzes what’s actually shown in that image: logos, people, environments, text overlays, and context clues. Teams can build categories that trigger based on visual signals, not just what the caption or article says. What can image categorization detect? Teams can build categories to detect and classify:

• Brand logos — tracking where and how often your brand appears in organic content
• Subject matter — identifying what an image is actually about (protests, products, events, people)
• Harmful or off-brand visuals — flagging imagery that poses reputational risk
• Context and environment — reading text overlays, scene settings, and situational signals
• Irrelevant content — filtering out mentions where images have nothing to do with your topic
• IP misuse — understanding when intellectual property is being misused online
• AI-generated content — flagging when an image has been AI generated

How is this different from what monitoring tools do today? Most monitoring tools index the text around an image — the caption, the headline, the alt text. They do not analyze what is inside the image itself. Smart Categories actually process the visual content, which means they surface signals that are entirely invisible to text-based monitoring. A brand logo appearing in fan-posted content will never show up in a keyword search. Image categorization catches it. What does video categorization do? Video categorization analyzes the first 30 seconds of a short-form video — both the visual content and the spoken audio — to classify mentions based on what is actually said and shown, not just the caption or hashtags. This is especially relevant for platforms where brand narratives and crisis content tend to spread fastest. Why focus on the first 30 seconds? The first 30 seconds typically determine whether a video gets watched, shared, or acted on. For crisis and brand monitoring, that window captures the hook, the claim, and the emotional framing that drives virality. It is also a practical scoping decision that keeps the feature fast and cost-effective at scale.

​

Setup and configuration

How do I turn on image categorization for a Smart Category? In the Smart Category settings, toggle on Mention Image under Data Fields. Once enabled, the AI will factor in the visual content of images when assigning that category. You can preview how the category classifies content before saving. Can I use image categorization alongside text categorization? Yes. You can configure a Smart Category to analyze any combination of content signals — post text, author data, and now image content. Combining signals generally produces more accurate and nuanced classification than using any single source alone. Is there a template for image-based categories? Logo detection is available as a point-and-click starting point. Teams can also build from scratch using a prompt-based approach, describing what visual signals they want the AI to look for. Can I preview image categorization results before publishing a category? Yes. The preview experience reflects all active data signals, including image content, giving teams real-time feedback on how their category performs against actual mentions before it goes live.

​

Use cases

What kinds of teams find image categorization most useful? Any team monitoring high-volume visual content benefits most. Specific use cases include:

• Brand and comms teams tracking logo visibility in earned and organic content
• Crisis teams flagging harmful, threatening, or off-brand visual content as it surfaces
• Government and public affairs teams monitoring visual narratives across social platforms
• Research and insights teams building richer datasets that include visual context
• Executive protection teams tracking executive perception across images
• Legal teams tracking IP misuse online
• Sponsorship teams quantifying how often a brand or logo appears in partner-created content

Can image categorization be used to track competitors? Yes. Teams can build categories to detect competitor logos or products appearing in content alongside their own brand — or to understand how competitors are being shown visually across platforms. What about synthetic or AI-generated images? Smart Categories can be configured to classify images based on visual signals consistent with AI-generated or synthetic media. Teams can build categories that flag content for further review based on those signals, which feeds directly into PeakMetrics’ broader synthetic content detection capabilities.

​

Availability

Is image categorization available now? Yes. Image categorization is generally available within the PeakMetrics platform today. Reach out to the PeakMetrics team at support@peakmetrics.com to get started. Is video categorization available now? Video categorization is currently available as part of an early access program. Contact your PeakMetrics team to discuss access and setup. Does image or video categorization cost extra? Visual content processing requires additional compute. Your account team can provide specifics on how image and video categorization are scoped within your plan.

​

Author data Smart Categories

​

The basics

What is author data in Smart Categories? Author data allows Smart Categories to factor in information about who posted a mention — not just what was posted. This includes account reach, follower counts, profile information, publisher details, and location signals. Teams can build intelligence models that reflect the actual influence and identity of the sources behind the conversation. What author-level signals are available? The current author data set includes:

• Account reach and follower count
• Profile information — bio, account type, handle, publisher details
• Location signals — where available from the author’s profile
• Engagement metrics — relative activity and influence signals
• Bot classification — whether a human or bot is pushing the narrative

Why does author data matter for categorization? A post from a sitting senator, a major journalist, and a bot network can all say the exact same thing. Without author context, they look identical. Author data lets teams build categories that reflect who is actually driving a conversation — so intelligence reflects reality, not just volume. How is this different from standard audience analytics? Standard analytics tell you aggregate demographic information about who’s engaging with content. Author data in Smart Categories goes further: it lets you build active classification rules around specific author characteristics, so mentions can be filtered, weighted, and analyzed based on the source, not just the content.

​

Setup and configuration

How do I enable author data in a Smart Category? In the Smart Category settings under Datasets Included, toggle on Author Data. The AI will then factor in author-level signals alongside any other data sources you have active when assigning that category. Can I build a category that classifies mentions based only on author data? Yes. Teams can configure a Smart Category to focus exclusively on author-level signals — for example, building a category that identifies posts from journalists, or filters to only show mentions from accounts with a follower count above a certain threshold. Can I combine author data with content and image signals? Yes, and this is often where the most powerful intelligence comes from. Combining who is speaking with what they are saying and what they are showing produces significantly richer and more accurate classification than any single signal alone. What is a dual-layer Smart Category approach? A dual-layer approach uses two separate Smart Categories that work together. One category identifies the right voices — for example, Beltway influencers, political journalists, or healthcare policymakers — based on author profile data. A second category then analyzes how those specific voices are expressing a sentiment or framing a topic. This approach allows teams to measure qualitative concepts like trust, credibility, or skepticism within a defined audience at quantitative scale.

​

Use cases

What can author data Smart Categories help teams do? Common use cases include:

• Separating real audience sentiment from bot-driven amplification
• Understanding how journalists talk about a brand versus how customers do
• Identifying whether a narrative is being driven by high-reach influencers or low-engagement accounts
• Building Beltway-specific intelligence for public affairs and government relations work
• Quantifying trust, credibility, and skepticism within a specific policy or industry audience
• Geo-filtering to understand whether backlash or support is regional, national, or concentrated in specific communities

Can author data be used to identify bot or inauthentic activity? Yes. Teams can build Smart Categories that classify mentions based on author signals consistent with bot-like or coordinated inauthentic behavior. This allows teams to separate manufactured amplification from genuine audience sentiment.

​

Availability

Is author data categorization available now? Yes. Author data is generally available as a dataset option within Smart Category configuration today. Are all author signals available for every source? Author data availability varies by platform and source. Reach and follower counts are broadly available across major social platforms. Location signals depend on what authors have made public in their profiles. Your PeakMetrics team can advise on signal availability for your specific monitoring sources.