This website uses cookies
Read our Privacy policy and Terms of use for more information.
Chrome Skill prompt: ABC Profile Scan
Objective: Analyze the social media profile or profiles provided for signals relevant to the ABC framework (Actors, Behavior, Content) used to assess influence operations and coordinated inauthentic behavior.Multi-tab execution: If this skill is run against multiple tabs simultaneously, produce a complete per-profile analysis (Steps 1–5) for each profile before any cross-profile assessment. Do not merge profiles — output each one separately, in full. After all individual analyses, add a Cross-Profile Summary section that identifies coordination signals across the set: shared narrative focus, similar account ages or structures, overlapping content or hashtag patterns, or other indicators of coordinated behavior.Step 1 — Identify the platformState which platform this profile is on: X, Facebook, Instagram, Threads, TikTok, Pinterest, YouTube, Reddit, LinkedIn, Bluesky, or Snapchat. If it is another platform, name it and apply the closest available approach. If this page is not a social media profile, stop and say so.Step 2 — Extract profile data (platform-specific)[X]- Handle and display name — note any mismatch between them- Bio text- Join date- Claimed location- Follower count and following count- Verification badge type: blue = paid subscriber, gold = organization, grey = government/official- Linked website- Profile photo — note if generic, illustrated, or possibly AI-generated (flag, do not assert)- Pinned post content and topic- Monetization icons near the DM button: pink icon = paid subscriptions enabled; black cash icon = tips enabled[Facebook — personal profile]- Name and profile URL/username- Bio or intro text- Listed location (current city, hometown)- Listed employer and education- Follower count and friend count (if visible)- Linked external websites- Blue Subscribe button visible = paid subscriptions enabled[Facebook — Page]- Page name and URL/username- Page category label- About text- Follower and like counts- Linked website- Blue Subscribe button visible = paid subscriptions enabled[Instagram]- Username and display name- Bio text and any link in bio- Post count, follower count, following count- Account type label if shown (personal, creator, or business — note the category)- Verified badge if present- Story highlight titles visible on the profile (do not open them)- Approximate posting frequency from visible grid- Crown icon on profile = paid subscriptions enabled[Threads]- Username and display name — note whether the Threads username matches the linked Instagram handle or differs (a mismatch may indicate rebranding)- Bio text and any link in bio- Follower count- Post count if shown- Verified badge if present- Account type label if shown (personal, creator, or business)- Titles or text of recent posts visible on the profile- Any linked Instagram account shown on the profile[TikTok]- Username and display name- Bio text- Any linked accounts shown (Instagram, YouTube, website)- Follower count, following count, total likes- Verified badge- Titles and descriptions of videos visible on the profile page- Any playlist or series names shown- Shopping bag icon on profile = TikTok Shop enabled[Pinterest]- Username, display name, bio, claimed location, and linked website- Follower and following counts; monthly views if shown- Board names and topics- Whether the account is personal or business[YouTube]- Channel name and handle- About/description text- Channel creation date — if not visible on the main channel page, look for a "More info" link near the channel description (the About tab has been removed from most YouTube channel layouts)- Subscriber count, total view count, video count- Listed country- Linked websites and social accounts listed in the channel description or visible info section- Verified badge- Titles and upload dates of recent videos- Playlist names and topics- Note whether ads are visible in videos or Super Chats appear in any listed live streams — both indicate active monetization- Thumbnail analysis for visible videos: * Do real people appear in thumbnails? If so, does the same person appear consistently across videos (suggesting an on-camera creator persona), or do faces vary, appear generic, or look possibly AI-generated? Flag — do not assert. * Does text overlaid on thumbnails match, expand on, or significantly diverge from the video title? Divergence — especially shocking or sensational thumbnail text not reflected in the title — is a clickbait signal. * Is the thumbnail visual style and template consistent across the channel's visible history, or is there a point where the style changes noticeably? A visual break may indicate repurposing or a change in production team. * Does thumbnail imagery use exaggerated facial expressions, fear-based visuals, or shock imagery beyond what the title suggests?[Reddit]- Username- Bio text if set- Account creation date- Post karma and comment karma — note the ratio and which is dominant- Subreddits visible in recent post history- Subreddits the user moderates, if shown- Topics of recent posts and comments[LinkedIn — personal profile]- Name and profile URL- Headline (current title and employer as stated)- About/bio text — extract what is visible; flag as possibly truncated if a "see more" inline link is present- Earliest visible entry in the Experience section — note the year and whether a "Show all [X] experiences" link appears (Gemini reads DOM labels, so it may report counts not visible on screen — trust the label as reported)- Education entries visible — note if a "Show all [X] education" link appears indicating additional entries are hidden- Join date if visible on the profile page (note: full details require "About this member" panel — see Step 6)- Connection count — note if under 500 despite claimed seniority- Any double first names (e.g., "Alex Alex") — a signal associated with AI-generated profile names- Verification badges shown and what was verified- Recent posts and comments visible on the profile — note if a "Show all" link appears indicating that the full activity history is not shown[LinkedIn — company page]- Company name and page URL- Industry, company size, and founded date- Follower count- Number of employees listed on the page- Recent posts visible[Bluesky]- Handle and display name — note whether the handle is a standard .bsky.social handle or a custom domain (e.g., @name.organization.org). A custom domain indicates the user has verified a link to an external website.- Bio/description text and any links- Follower count and following count- Post count- Join date if visible- Any labels visible on the profile — labels are community-applied annotations that may indicate the account has been flagged or categorized by moderation labelers- Pinned post content if present- Profile photo — note if generic, illustrated, or possibly AI-generated (flag, do not assert)[Snapchat — public/creator profile]- Username, display name, bio, subscriber count, and any links to other platforms- Titles or topics of any public stories or Spotlight videos visible- Note: most Snapchat content is ephemeral; report only what is directly visibleStep 3 — Map to ABC frameworkA — Actor signalsAssess what is visible about the identity and origin of the account:- Is the identity claim specific and verifiable, or generic and vague?- Does account age fit the apparent follower volume and activity level?- Are there mismatches between bio claims and other visible signals (e.g., claimed location vs. content focus, claimed profession vs. posting topics)?- Does the profile image appear to be a real person, stock photo, illustration, or possibly AI-generated? Flag if so — do not assert.- YouTube only: flag if thumbnail faces are inconsistent across videos or a break in visual style suggests channel repurposing.- Reddit only: flag if account age and karma suggest a recently created or low-history account.- LinkedIn only: flag if claimed experience predates the join date or connection count is inconsistent with claimed seniority; if a "Show all [X] experiences" link is visible, note the full history may extend further back.- Note any visible monetization indicators (subscription icons, shopping integrations, tips features) — active platform monetization indicates a financial relationship with the platform and requires meeting minimum eligibility thresholds for followers and content activity.B — Behavior signals (visible only)Only report what is directly observable on the page. This skill does not scroll or trigger additional content loading — behavior analysis is limited to what is currently visible. Flag to the user if the visible sample appears small or if pagination links suggest more content exists.- What proportion of visible content is original vs. reshared, reposted, or retweeted?- Are replies or comments to other accounts visible, or is the account largely broadcasting without interaction?- Does visible content suggest consistent posting, bursts, or sparse activity?- Platform-specific: * X: ratio of retweets to original posts; presence or absence of replies * Reddit: whether karma is dominated by posts or comments; which subreddits are active * YouTube: upload pace and consistency from visible video history * TikTok: presence of duets or stitches in visible videos * Pinterest: ratio of repins to original pins * Instagram: posting frequency and whether highlights suggest sustained or recently built-up presenceC — Content signalsNote: content analysis is limited to what is currently loaded on the page. This skill does not scroll or auto-paginate. If the visible sample is small, flag this and recommend the user scroll to load more before re-running.- Primary topics and narratives in visible posts, videos, or pins- Dominant emotional register: outrage, fear, enthusiasm, mockery, inspiration- Recurring hashtags, phrases, or keywords- Entities being promoted: people, movements, products, organizations- Entities being criticized or attacked- Language quality — does it read as natural, stilted, or possibly translated?- Any visible claims that appear factually contestable- YouTube only: whether thumbnail text diverges from video titles in ways that suggest clickbait framing — note specific examples where the gap is significantStep 4 — Handling exceptionsIf the page requires a login to view content, or the profile is private or restricted, note what was and was not accessible. Only report what is visible — do not infer or fabricate signals. If a section has no visible data, omit it.Step 5 — Final response- State the platform on the first line- Present findings under three headings: Actor, Behavior, Content- Use sentence case throughout- Flag signals that warrant further investigation; do not conclude inauthenticity- If signals appear across more than one vector — for example, a recently created account whose content closely mirrors a known narrative cluster — note the combination explicitly- Close with a one-line summary: the strongest signal observed and which vector it belongs to- After the summary, always include: "This analysis covers only what was loaded when the skill ran. Scroll the full page and re-run for a more complete picture."- When running across multiple tabs, clearly label each profile's output block with the platform and handle (e.g., "Profile 1: X @username") so outputs are distinguishable when read togetherStep 6 — Next stepsFor every platform, always include these steps regardless of which vector was strongest:- "Search this account's username on whatsmyname.app to check for the same handle registered on other platforms."- "If you have the My OSINT Training bookmarklets installed (tools.myosint.training), run the relevant bookmarklet for this platform. The bookmarklets extract numeric account IDs that persist across username changes — permanent identifiers for tracking an account over time. Available for Bluesky, Facebook, Instagram, Snapchat, Threads, TikTok, and Venmo; additional bookmarklets retrieve profile photos, locked content, and country-of-origin data. All run client-side."Based on which ABC vector showed the strongest signals, suggest 3 specific investigative actions from the platform-appropriate list below. Lead with the actions most relevant to what was found. Frame each as a specific click, navigation step, external tool, or follow-up question. Do not list all options — select and prioritize.[X]- Actor: "Click on the join date shown on the profile to open the 'About this account' panel. This shows the account creation date, the country of operation, and the country of the phone number used to verify the account. It is one of the most useful transparency features on X and is not visible in the main profile view."- Behavior: "Navigate to the Replies tab. Accounts that broadcast without replying show a distinct behavioral pattern — note whether replies are present or absent."- Content: "Search this handle in X Advanced Search to filter posts by date range or keyword — useful for checking whether content focus shifted at a specific point in time."[Facebook — Page]- Actor: "Click on the page name to view additional details, then select 'Transparency and privacy policy.' The expanded view shows the creation date, any prior name changes, and the confirmed country of origin for the managers."- Actor/Content: "Visit the Meta Ad Library at facebook.com/ads/library, search this page name, and run this skill on the results to extract active and past ad content, spend ranges, and audience targeting."- Actor: "Paste this page's URL into monetization.wtf to check Meta monetization program enrollment history, including the first enrollment date and any demonetization gaps."[Facebook — personal profile]- Actor: "Click on the profile name to view additional details about location, work, and education."- Actor: "If the profile is public, check the Photos tab to see whether profile photos have changed significantly over time."- Actor: "Paste this profile's URL into monetization.wtf to check whether this account is enrolled in any Meta monetization programs."[Instagram]- Actor: "Navigate to the Tagged tab to see content others have tagged this account in — useful for cross-referencing claimed identity against external documentation."- Actor/Content: "Check whether this account appears in the Meta Ad Library at facebook.com/ads/library — it covers Instagram as well as Facebook."- Actor: "Paste this account's URL into monetization.wtf to check Meta monetization program enrollment — monetization.wtf covers Instagram as well as Facebook."[Threads]- Actor: "Navigate to the linked Instagram profile — Threads accounts are tied to Instagram accounts, and the Instagram profile may contain additional signals including follower history, tagged photos, and story highlights. Run this skill on the Instagram profile."- Actor/Content: "Check the Meta Ad Library at facebook.com/ads/library — it covers Threads as well as Facebook and Instagram."- Actor: "Paste this account's URL into monetization.wtf to check Meta monetization program enrollment history."[TikTok]- Actor: "If other accounts are linked in the bio, navigate to those profiles and run this skill there."- Actor/Content: "Check TikTok's Ad Library at library.tiktok.com for any paid content from this account."- Content: "Sort videos by oldest to see the account's earliest content and whether topic focus or production quality shifted significantly."[Pinterest]- Actor: "If a website is linked in the profile, run this skill on that page to check for corporate or ownership signals."- Content: "Click into the most active boards to see full pin history and identify which external domains are linked to consistently."- Behavior: "Check whether boards were created in bursts or gradually over time — board creation dates are visible when you open each board."[YouTube]- Actor: "Look for a 'More info' link near the channel description — this shows the creation date, listed country, and contact email."- Actor: "Search the channel name on Fameswap (fameswap.com) and Swapd (swapd.co) to check whether it has been listed for sale — these are the primary marketplaces where YouTube channels are bought and sold."- Behavior: "Sort videos by Oldest using the filter on the Videos tab to see what the channel was posting when it first launched. Look for shifts in language, topic, or production style that may indicate a channel was repurposed after sale."[Reddit]- Behavior: "Sort this user's history by New to see current activity, and by Top to identify highest-engagement content. Note whether topics are consistent or have shifted."- Behavior: "Click through to the subreddits where this account posts most frequently — assess whether those communities are relevant to your investigation."- Behavior: "If comparing multiple accounts, check for overlapping subreddit activity — a coordination signal worth documenting."[LinkedIn — personal profile]- Actor: "Click 'Contact info' to check for linked websites, email addresses, or phone numbers not visible in the main profile view."- Actor: "Click the profile name to open the 'About this member' panel — this shows join date, when the profile photo was last updated, and verification details."- Actor: "Before comparing claimed experience to work history, check whether a 'Show all [X] experiences' link appears at the bottom of the Experience section. If so, click it to load the full work history — earlier entries may not be visible until you do. Then compare the earliest entry against the LinkedIn join date from 'About this member.'"[LinkedIn — company page]- Actor: "Click the People tab and note whether employees show as named accounts or as 'LinkedIn Member' — the latter indicates fake employee accounts that have been removed by LinkedIn."- Actor: "Click the Insights tab and look for sudden spikes in follower counts, which can result from fake job applications inflating the follower number."- Actor: "Check the Jobs tab for unusually high job posting volume relative to company size — posting many jobs is a method used to build fake employee networks."[Bluesky]- Actor: "If the handle uses a custom domain rather than .bsky.social, navigate to that domain — the site may reveal organizational affiliation or contact details."- Actor: "Check for any labels displayed on the profile — community-applied annotations that may flag behavioral patterns or affiliations not in the bio."- Behavior: "Search for this handle in Bluesky Starter Packs at bsky.app — clustered packs can reveal which communities or networks the account is associated with."[Snapchat — public/creator profile]- Actor: "Snapchat usernames cannot be changed once set, making them a stable cross-platform identifier. Search the username on whatsmyname.app to check for matching accounts on other platforms."- Content: "Open any visible public stories or Spotlight videos to check for location signals, real-world identity clues, or linked platform references not visible from the profile page alone."- Actor: "Snapchat has no transparency panel comparable to Facebook or X. Focus attribution work on cross-platform username matching and any external links or contact information visible on the profile."
Chrome Skill prompt: Entity Extractor
Paste the prompt below into a customer Skill in Chrome. Read more here.
Objective: Extract all named entities and contact information visible on this page or pages.Multi-tab execution: If this skill is run against multiple tabs simultaneously, produce a complete per-page extraction (Sections 1–4) for each page, clearly labeled with the page title and URL. After all individual extractions, add a Cross-Page Summary flagging any entity — person, organization, or contact identifier — that appears on more than one page. The same name or identifier turning up across multiple pages is a relationship or coordination signal worth isolating.1. People - Full names mentioned - The section of the page where each person appears (e.g., Team, Angel Investors, Advisors, Board) — this is their relationship to the organization - Their description as given on the page (title, role, or any qualifying phrase) - All hyperlinks attached to or associated with each person's name or listing. Extract every link found — do not stop at the first one. For each person, separately record: * Twitter/X URL (if present) * LinkedIn URL (if present) * Any other URLs (personal site, GitHub, etc.) - URL stripping rule: if any URL matches the pattern google.com/search?q=https://... strip the wrapper and record only the actual destination from the q= parameter. Apply this to every URL extracted in this section.2. Organizations - Company or organization names - Any associated addresses, registration numbers, or jurisdictions3. Contact and digital identifiers - Email addresses - Phone numbers - URL extraction rule: before reporting any URL, check whether it matches the pattern google.com/search?q=https://... — Chrome wraps some link destinations this way. If it does, strip the wrapper and report only the actual destination URL from the q= parameter. Never output a google.com/search URL as a result. - Social media links: for each one, report the full destination URL (e.g., twitter.com/username, linkedin.com/in/name, instagram.com/handle) — not just the platform name. Check underlying link destinations, not only visible text. Icon-only links still have href values — extract those. - Domain names or websites linked from the page (full URLs)4. Dates and locations - Any specific dates mentioned (founding, events, filings) - Physical addresses or geographic locations5. Handling exceptions If the page has limited content or the content isn't compatible with this request: - Extract what is available and note what's missing - Flag if the page requires a login or is behind a paywall6. Final response Present results as clean labeled lists. Use sentence case. Don't interpret or editorialize — just extract. If a field has no data, omit it rather than writing "none found."After presenting results, ask: "Would you like any section reformatted as a table for pasting into a spreadsheet? For example: a People table, an Organizations table, or a Contact identifiers table."When generating a People table, use these exact columns — one per column, nothing collapsed or combined: Name | Page description | Relationship to company | Twitter/X URL | LinkedIn URL | Other URLsLeave a cell blank if no value was found for that column. Do not merge link types into a single "hyperlinks" column.If you identified relationships between any of the people or organizations listed, offer to map those connections.
Chrome Skill prompt: Image Analysis
Paste the prompt below into a customer Skill in Chrome. Read more here.
Objective: Identify and visually assess all images on this page or pages for signals of AI generation, metadata availability, and contextual consistency. This skill performs visual triage only — it cannot extract EXIF metadata or perform SynthID watermark detection, both of which require external tools listed in Step 5.Multi-tab execution: If this skill is run against multiple tabs simultaneously, produce a complete per-page analysis (Steps 1–4) for each page, grouped under a clear page label showing the page title and URL. After all individual analyses, add a Cross-Page Summary noting any image src URL that appears on more than one page — the same image hosted at the same URL across multiple sites is a reuse or syndication signal worth flagging.Step 1 — Inventory all visible imagesFor each image visible on the page, record:- Source URL (full URL, not a relative path — if relative, note the page domain)- File name: extract the filename from the end of the src URL and examine it for any of the following: * AI tool names embedded in the filename: gemini, dalle, dall-e, midjourney, firefly, stablediffusion, sora, imagen, kling, runway, flux, or similar — these directly indicate generation source * Stock photo identifiers: patterns like shutterstock, getty, istock, depositphotos, or numeric IDs matching stock library conventions * Camera-originated patterns: IMG_, DSC_, DCIM, or similar prefixes typical of device-generated filenames suggest a real camera capture * Prompt or description text: some AI tools generate filenames from the prompt used — a filename like "nurse-crying-hospital-ward.jpg" on an image purporting to be documentary photography is worth flagging * Geographic or organizational terms that may indicate intended use, target audience, or origin * Hash-like strings (long random alphanumeric sequences) — common for AI-generated images and CDN-hosted content * Sequential numbering patterns suggesting bulk generation (e.g., image_001.jpg through image_099.jpg) * Dates embedded in the filename that contradict the claimed publication date or depicted events Flag any filename that contains notable terms and note what they suggest.- Alt text (if present)- Caption or credit line (if present in surrounding HTML)- Context: which section of the page the image appears in, and any headline or body text immediately adjacent to itIf images have not fully loaded, or the page is image-heavy and only a portion is in the viewport, flag that the inventory may be incomplete.Step 2 — Visual AI-generation assessmentFor each image, work through the following checks in order. Note every specific signal observed — do not only report a rating.Anatomical and subject assessment:- Overall perfection check: does this person look too polished or groomed for the situation depicted? Magazine-quality aesthetics in a crisis, conflict, or candid scene is a primary red flag.- Skin texture: at high zoom, does skin show natural pores, minor imperfections, and asymmetries? Plastic-looking, overly smooth, or patchily detailed skin is a signal. AI-smoothed skin is especially common on faces and necks.- Eyes: check that light reflections are consistent in both eyes; pupil shape and size should match. Inconsistency suggests composite generation.- Ears: asymmetry, missing cartilage detail, or merging with hair.- Teeth: unnaturally uniform or too perfect; look for absence of natural imperfections.- Hair: individual strands should be visible. A painted, blended, or uniformly rendered appearance suggests AI generation.- Fabric and clothing: patterns should align correctly across folds and seams; natural wear and wrinkles should be present. AI often produces fabric that looks draped rather than worn.- Accessories: jewelry, glasses, and watches often appear merged with skin or rendered flat rather than three-dimensional.- Note: AI image quality has improved significantly. Extra fingers or fused digits were once reliable tells but are now less common in images from major models. Anatomical anomalies are worth flagging but their absence no longer indicates authenticity.Physics and geometry:- Shadow consistency: identify the primary light source (brightest highlights). All shadows should point away from it in the same direction. Multiple shadow directions in a single-source scene (e.g., outdoor sunlight) is a physics violation AI frequently produces.- Vanishing point: parallel lines on buildings — rooflines, window rows — should converge at a single point on the horizon. If rooflines point left while window lines point right on the same structure, this is an AI assembly error.- Reflections: objects reflected in water, glass, or mirrors should connect to their reflections at right angles to the reflective surface. Impossible reflection positions or misaligned reflected images are geometric failures.- Depth of field: background blur should match what a real lens at that distance and aperture would produce. Artificial bokeh or uniform sharpness across implausible depth ranges is a signal.Texture and surface analysis:- Surfaces that should be imperfect — walls, concrete, water, old fabric — often appear AI-smoothed or airbrushed. Look for uniform smoothness where natural variation should exist.- Text within the image: AI text rendering has improved but still fails on complex signage, protest signs, or overlapping text. Note garbled letters, floating characters, or inconsistent fonts.- Background warping or distortion near the edges of subjects, particularly in areas where the subject meets the background.Temporal and contextual logic:- Season and weather: does vegetation, clothing, and lighting match the purported date and location? Leafless trees in claimed summer footage, or winter clothing in bright tropical light, are contextual failures.- Technology and era: do visible devices, vehicles, infrastructure, and clothing styles match the claimed time period? Anachronisms suggest image generation or manipulation.- Geographic consistency: does the visible architecture, signage, and landscape match the claimed location? AI assembles visual elements without necessarily understanding their geographic specificity.- Caption and credit check: does the image content match the caption or surrounding text? Note any contradictions between what is depicted and what is claimed.Crowd and behavioral patterns (flag only if a crowd scene is present):- Are there too many people of similar age, appearance, or clothing style? Real crowds have demographic diversity; AI crowds often show artificial uniformity.- Do people's attention patterns look natural, or are too many looking in the same direction or directly at the camera?- Do facial expressions and body language match the emotional intensity of the supposed event?- Does personal spacing and body language look like natural human behavior?Intuition and production paradox:- Does the image feel "too good" for its claimed origin — cinematic quality from an amateur source, or perfect documentation of a chaotic event?- Is the content designed to trigger an emotional response (fear, outrage, sympathy) in ways that seem calibrated rather than organic?- Flag images that produce an uncanny-valley feeling even when no specific technical flaw is identifiable — this intuitive signal has investigative value.Assign each image one of these assessments:- No visible signals- Low concern (one minor signal, uncertain)- Moderate concern (two or more signals, or one strong signal)- High concern (multiple clear signals across more than one category)Step 3 — Contextual consistency checkFor each image:- Does the image content match the caption, headline, or surrounding text?- If the text makes a date or era claim (e.g., "photo from 1985"), does the image content match — clothing, vehicles, technology, signage, image quality?- Does the stated photographer credit or source agency match the image style or apparent origin?- Does the image style fit the publication's or site's typical visual presentation, or does it appear visually inconsistent with the rest of the page?Step 4 — Handling exceptions- This skill only sees images loaded in the current viewport. On long articles, pages with embedded screenshots, or infinite-scroll feeds, most images will not be visible. Always report how many images were found and note explicitly if the page appears to contain more content below the fold. Instruct the user to scroll the full page to load additional images, then re-run the skill.- If an image requires interaction to load (carousel, gallery, lightbox), note it as not assessed and flag it for manual review.- If an image loads as a CSS background rather than an HTML img element, note that its URL may not be directly extractable via this skill.Step 5 — Final responseFor each image, output a vertical block in this exact format. Do not use a horizontal table. Do not use Markdown heading syntax (no #, ##, or ###) anywhere in the output — use only bold (**text**) for labels. Every field must be on its own line with a hard line break after it.When running across multiple tabs, open each page's set of image blocks with a bold page label on its own line — **Page: [page title] — [URL]** — before the first image block for that page. Complete all image blocks for one page before moving to the next.---**Image [number]: [brief descriptive label]****Src URL:** [full URL]**File name:** [filename extracted from URL, plus any flags]**Alt text:** [value or "none"]**Caption / credit:** [value or "none"]**AI assessment:** [No visible signals / Low concern / Moderate concern / High concern]**Signals:** [bullet list of specific signals observed, or "none"]**Contextual flags:** [any contradictions between image and surrounding content, or "none"]---Repeat this block for every image. After the final block, output a plain text list of all src URLs, one per line, for copy-paste into external tools.Then include this block verbatim:"## Next steps for flagged images- EXIF/IPTC metadata: [exifviewer.info](https://exifviewer.info) or [exifdata.com](https://exifdata.com) — most platforms strip EXIF on upload; an empty result means stripping, not a clean file. Google AI images carry IPTC Credit: 'Made with Google AI' — check with [InVID](https://www.invid-project.eu/tools-and-services/invid-verification-plugin/).- Reverse image search: [Google Images](https://images.google.com), [TinEye](https://tineye.com), [Yandex Images](https://yandex.com/images) (best for faces).- AI image detection: [Image Whisperer](https://imagewhisperer.org).- Forgery/pixel analysis: [Image Verification Assistant](https://mever.iti.gr/forensics/) (forgery probability score) · [Forensically](https://29a.ch/photo-forensics/) (noise and frequency analysis).- Deepfake detection: [TrueMedia.org](https://www.truemedia.org).- SynthID only detects watermarks on Google-generated images — a negative result does not rule out AI generation."Only if one or more image src URLs directly contain the string "wp-content/uploads/" — meaning the images themselves are hosted on a WordPress installation, not merely that the page discusses WordPress — also include this block:"## WordPress investigation steps- The image URL may include a resized version (e.g., '-300x300' before the file extension). Remove the size suffix to access the original upload, which is more likely to retain EXIF or IPTC metadata. Example: 'image-300x300.jpg' → 'image.jpg'.- To enumerate all media uploaded to this WordPress site, navigate to: [base domain]/wp-json/wp/v2/media — this lists image names and URLs including files attached to unpublished posts.- To see all author accounts: [base domain]/wp-json/wp/v2/users?per_page=100 — usernames often appear on other platforms and can be searched via whatsmyname.app.- Consider using a VPN or Tor when accessing the WordPress API directly, as site administrators can see API access in their logs.- Run the domain through HackerTarget's WordPress Security Scan (hackertarget.com/wordpress-security-scan/) to identify the theme, plugins, and additional author accounts. Unique plugin combinations can link multiple sites run by the same operator."
