ImagesiftBot is Hive's crawler for publicly available images. It also saves page text that sits with those images. Hive stores both in an index used for similar-image search. You name ImagesiftBot in robots.txt when you want to change that access.
What the crawler saves
Hive's crawler page says ImagesiftBot scrapes public images for Hive's web intelligence products. Product naming on that page is Imagesift by Hive.
Each saved image includes the host URL and the text on that page. Alt text associated with the image is saved too. Hive says ImageSift analyzes the downloaded images and the page text. That information is stored in an index. Hive's products use the index to search for similar images and to retrieve them.
The ImageSift homepage says you can identify duplicates and variants of an image across the web. Visitors get a reverse image search form. Upload a file, or paste an image link, then run Search. Uploads accept JPEG, PNG, TIFF, GIF, WebP, and AVIF. Placeholder text in the link box reads "Paste Image Link Here".
Nothing on the about page describes an AI training dataset. Aiola's ImagesiftBot entry sets the category to indexing. Its notes describe public images and page text in an index for similar-image search and retrieval. The shared Indexing bucket is described as crawled for AI or search indexing. That description covers the bucket, not a separate Hive claim. Notes on the record are the statement for this bot.
Match the published user agent
Hive publishes this user-agent string:
Mozilla/5.0 (compatible; ImagesiftBot; +imagesift.com)
Inside that string, the token is ImagesiftBot. Aiola stores that same sample on the crawler entry. Its robots.txt token is ImagesiftBot. Vendor on the record is Hive, and the vendor domain is thehive.ai. Last verification on the record is 2026-10-23.
Hive's heading writes ImageSiftBot, with a capital S. Both the user-agent and the robots.txt samples use ImagesiftBot. Put ImagesiftBot in your rule so it matches those samples.
IP range type on the entry is none. No fetched ranges are listed for this bot. No IP verification is published for this bot, so treat the token as a claim.
Block every path
Hive respects standard robots.txt directives that target ImagesiftBot. Use this group when you want every path disallowed:
User-agent: ImagesiftBot
Disallow: /Serve the file at /robots.txt on the hostname you care about. Aiola's robots.txt guide says to do that for each relevant hostname. After you deploy, open the public robots.txt URL and read the response.
Do not assume a named group inherits path rules from the wildcard group. If an ImagesiftBot group is already present, edit that group. Contradictory duplicate groups should be removed, the guide says.
Sample files in the guide place this token with other indexing bots. A sample can include a Content-Signal line. Hive's about page does not mention Content-Signal. Those samples are not proof that every token honors the lines, the guide says.
Limit paths or follow Googlebot
Hive's own sample allows the site and blocks one prefix:
User-agent: ImagesiftBot
Allow: /
Disallow: /private/Hive says that sample lets the crawler fetch all pages except paths under /private/.
Crawl-delay is supported in the same kind of group. Count that number as the minimum seconds between the start of consecutive requests. Hive's example sets the value to 5. From there, the crawler splits the day into five-second intervals. It issues at most one request to your domain inside each interval.
If no rule targets ImagesiftBot, and a rule targets Googlebot, the Googlebot rule is used. Hive shows a wildcard block beside a Googlebot exception:
User-agent: *
Disallow: /
User-agent: Googlebot
Allow: /
Disallow: /private/In that example, ImagesiftBot fetches all pages except those under /private/. A wildcard Disallow does not replace that Googlebot group. Hive states this fallback only when a Googlebot rule exists. It does not state what happens when the file names neither agent.
Directory notes say the same two behaviors in shorter form. ImagesiftBot robots.txt rules are honored, those notes say. Googlebot directives apply when no ImagesiftBot group exists.
What Aiola can show you
Aiola Analytics names 70 crawlers, including ImagesiftBot. Hosted Analytics needs Pro or the 7-day trial.
Only a first-time Pro Monthly customer can start that trial. That customer must have no prior Stripe subscription. Included usage is 100,000 analytics events, priced at $24 per month. Checkout collects a card. If Stripe history cannot be checked, no trial is granted. An existing active, trialing, or past-due subscription opens the billing portal. Pro Yearly has no trial. Yearly billing at 100,000 events costs $240. GUI Lifetime has no trial and does not include hosted analytics.
Crawlers do not execute JavaScript, so the browser tag cannot see them. Set Up Analytics says to add one server or edge snippet. Put it on the stack that receives page requests. Posts from that snippet go straight to https://aiola.app/api/bot-crawl-ingest. Ad blockers do not apply to that server post.
ImagesiftBot is in the snippet's user-agent test. A match sends tracking_id, url_path, url_host, and ua. Country is added when the platform provides it. Status is sent where the stack can observe it. Analytics docs say only known crawler user agents are stored. Hosts must match the configured domain.
Skipped path endings include png, jpg, jpeg, gif, webp, avif, and svg. Also skipped are ico, css, js, mjs, and map. Skipped fonts are woff, woff2, ttf, otf, and eot. JSON, webmanifest, mp4, webm, and pdf are skipped too. TIFF and HTML are not on that list. Requests for robots.txt, llms.txt, and sitemap files are kept.
Next.js not-found handling in the guide does not apply that asset skip. It can post a crawler 404 when the user agent matches.
Body size stops at 10,000 characters. Path size stops at 2,000 characters, and host size stops at 500. User-agent text stops at 1,000 characters. Traffic is capped at 6,000 posts per minute per site per isolate. Country must be two letters when it is sent. Status values from 100 to 599 are kept. Any other status is stored as 200.
Without Pro or an active trial, the bot collector returns 402. It does not store that event.
AI view totals include crawls, bots, pages, and a timeseries. Four buckets appear there too. Search and index kinds map to Indexing. This bot's kind is index, so a stored hit sits in Indexing. Crawler 404s and discovery-file health appear in that view. Visibility runs from 0 to 100 and includes robots.txt, llms.txt, and sitemap health. Those inputs do not score whether ImagesiftBot followed your file.
Contact Hive
Questions about ImagesiftBot can go to [email protected]. Hive lists that address for an opt-out of crawling as well.
Try it in Aiola: Download Aiola for free. Desktop installs run on macOS 12+ and Windows 10+. Linux is not supported. New crawler posts are stored only with Pro or an active trial.
