Which sources do Kagi search results and Kagi Assistant cite?
Kagi search combines its own index with other providers' search results. Kagi's official page (help.kagi.com, original text read on 2026-11-12, the page has no date) states:
| Source | What the official text says |
|---|---|
| Own index | The web index is internally called Teclis, the news index TinyGem |
| Search result providers | Anonymous API calls to “all major search result providers worldwide”, plus specialized engines such as Marginalia; the page does not name Yandex, Brave or Mojeek |
| Other data | The page names Wolfram Alpha, Apple, Wikipedia, Open Meteo, Yelp and TripAdvisor |
| Kagi Assistant | Web access is provided by Kagi Search and can be toggled with a button; it can fetch URLs pasted into the conversation, up to 50 MB |
Neither official page says how a site owner gets indexed or submits a site, or how Assistant citations are displayed. For both, we write “the vendor doesn't say”.
What is Kagibot, and how do I write robots.txt for it?
Kagibot is the crawler for Kagi search. Official page kagi.com/bot (original text read on 2026-11-12, the page has no date):
| Item | Official statement |
|---|---|
| User-Agent | Mozilla/5.0 (compatible; Kagibot/1.0; +https://kagi.com/bot) |
| robots.txt | Standard rules that target Kagibot are honored |
| When there is no Kagibot rule | If there is no rule for Kagibot but there is one for Googlebot, Kagibot follows the Googlebot rules; in the official example, the * group has Disallow: / and the Googlebot group has Allow: / with /private/ disallowed, so Kagibot crawls everything except /private/ |
| Source IPs | 4 are listed, with hostnames under kagibot.org (amd, donna, knet, ogi); the page does not say how to verify them |
| Contact | support@kagi.com |
# Let Kagi's crawler in; keep private paths out User-agent: Kagibot Allow: / Disallow: /admin/ User-agent: * Allow: / Disallow: /admin/ Disallow: /api/ Sitemap: https://your-site.com/sitemap.xml
This is a sample: give Kagibot an explicit allow group and keep private paths blocked as before. For the general approach see How do you write robots.txt for AI crawlers?; for crawler names see AI crawler user agents: a side-by-side table.
Are our own four sites open to Kagibot?
On 2026-11-12 we fetched the robots.txt of four sites (all 200) and checked the home pages for Kagibot with Python's robotparser:
| Site | robots.txt | For Kagibot |
|---|---|---|
| mochistamp.com | 310 bytes; the * group disallows /api/ (allowing /api/content.js and /api/strings/), /files/ and /admin.html; has a Sitemap | Home page allowed |
| gammaslide.com | 608 bytes; the * group disallows /d/, /decks, /account and /admin; a comment says AI crawlers are welcome; has a Sitemap | Home page allowed |
| kagentx.com | 174 bytes; the * group has Allow: / and disallows /admin.html, /candidates.html and /api/; has a Sitemap | Home page allowed |
| pagent.kagentx.com | 156 bytes; the * group disallows /i18n/; has a Sitemap and an llms.txt note | Home page allowed |
None of the four sites names Kagibot or has a Googlebot-specific group, so Kagibot falls under the * group. This is computed from the rule text, not from real visits; whether Kagibot appears in the server logs is unverified, and how to check logs is in How do I check whether GPTBot and other AI crawlers have visited my site?.
How do I make Kagi more likely to recommend you, step by step?
1. Allow Kagibot in robots.txt
Do not block them by name without a specific reason. If you have a Googlebot-specific rule, remember that Kagibot follows it.
2. Be findable in other search engines too
Kagi's official page says it calls other search result providers, so being found in other search engines may also let Kagi find you (an inference; the vendor doesn't say). For Bing and Brave, see How to get Brave Search to index and recommend my site and How do I get DuckDuckGo's AI answers to recommend me? Duck.ai and Search Assist.
3. Server-render the page body and state what it does and who it is for
Whether Kagibot runs scripts is something the vendor doesn't say, so don't rely on script rendering.
4. Check the logs to confirm they actually visited
Look in the logs for the Kagibot User-Agent and check whether the source IP's hostname is under kagibot.org.
5. Test with the same set of questions
Turn on web access in Kagi Assistant and ask "which product of this kind is good"; ask each question 3 times and note whether you are mentioned and which URLs are cited; for how to record it, see How do you measure AI visibility? Four metrics and a table you can follow.
Related guides: How do I get Perplexity to recommend my product?, How to get Brave Search to index and recommend my site, How do I get DuckDuckGo's AI answers to recommend me? Duck.ai and Search Assist.
FAQ
Where do Kagi search results come from?
Kagi's official page says: from its own web index Teclis and news index TinyGem, plus anonymous API calls to the major search result providers worldwide and specialized engines such as Marginalia; also Wolfram Alpha, Apple, Wikipedia, Open Meteo, Yelp, TripAdvisor and others. The page does not name Yandex, Brave or Mojeek.
What is Kagi's crawler called?
It is called Kagibot. The official page kagi.com/bot gives the User-Agent Mozilla/5.0 (compatible; Kagibot/1.0; +https://kagi.com/bot) and lists 4 source IPs with hostnames under kagibot.org. The page does not say how to verify that a visitor is genuine.
How does robots.txt apply to Kagibot?
The official page says standard rules that target Kagibot are honored; if there is no rule for Kagibot but there is one for Googlebot, Kagibot follows the Googlebot rules. So do not write rules for Googlebot that block by mistake.
Where does Kagi Assistant get its web information?
The official page says web access is provided by Kagi Search and can be toggled; it can also fetch URLs pasted into the conversation, up to 50 MB. The page does not say how citations are displayed or how a site owner gets indexed; the vendor doesn't say.
How do I submit a site to Kagi for indexing?
Neither of the two official documentation pages we read, nor kagi.com/bot, describes how to submit a site; the vendor doesn't say. What you can do is keep robots.txt open to Kagibot and make the site findable through other search sources (the search result providers that Kagi's official page says it calls).
What do the robots.txt files of our four sites say about Kagibot?
Tested on 2026-11-12: the robots.txt of mochistamp.com, gammaslide.com, kagentx.com and pagent.kagentx.com does not name Kagibot and has no Googlebot-specific group; by the rule text, Kagibot is allowed on all four home pages. Whether Kagibot appears in the server logs is unverified.
Does Kagi have many users, and is it worth doing?
We have no data on Kagi's user numbers or referral traffic; unverified. The cost of what you can do is low: make sure robots.txt doesn't block it by mistake, and the rest is the same work as for Google and Bing.
Sources
- Kagi · Search sources (Kagi documentation) (Read in full 2026-11-12, page has no date)
- Kagi · Kagi Assistant (Kagi documentation) (Original text read on 2026-11-12; only the model list is dated 2026-07-01)
- Kagi · About Kagibot (Read in full 2026-11-12, page has no date)
- MochiStamp · robots.txt of mochistamp.com, gammaslide.com, kagentx.com and pagent.kagentx.com (Read 2026-11-12)
- kagentx evaluations: the data on the case study pages; each example gives its date.