Applebot and Siri AI: What iOS 27 Changes for Brands
Siri AI answers from the web in iOS 27. What Apple says about Applebot and source links, what is still unknown, and how to test Siri in your category.
On this page
Applebot is the crawler that feeds Siri AI's web answers, and it is the one part of Siri visibility you control today. Apple shipped Siri AI on September 14, 2026, as an English beta that answers questions from the web. Apple has published no citation rules, so start by making sure Applebot can reach the pages you want used.
Everything else about how Siri picks and credits sources is still undocumented. This guide separates what Apple has said from what it has not, then gives you two checks you can run this week: one on your crawler access, one on Siri's answers in your category.
What changed for Siri in iOS 27?
Siri became an answer engine. Since iOS 27, it can look up current information on the web and write an answer, instead of handing you a list of search results. Apple calls the feature Siri AI and still labels it a beta.
Apple's launch announcement and its software release post, both dated September 14, 2026, set out the terms:
| Question | What Apple says |
|---|---|
| Release | September 14, 2026, rolling out in beta |
| Platforms | iOS 27, iPadOS 27, macOS 27, watchOS 27 and visionOS 27 |
| Language | English first; French, Japanese, Korean, Portuguese and Spanish due in October |
| Where it is missing | Not in the EU at first on iOS, iPadOS and watchOS (Mac and Vision Pro are included); not in China |
| Who can use it | Apple Intelligence devices, such as iPhone 16 or later and iPhone 15 Pro; not for users under 13 |
| Models | New Apple Foundation Models, which Apple says it built in collaboration with Google and its Gemini models, running on device and on Private Cloud Compute |
| Limits | Server-side features have daily usage limits; Apple says expanded access will cost a fee in the future |
Two details matter for brands. First, Siri AI answers broad questions from the web, and Apple's own example is finding the best local coffee spot in a new city. That is a recommendation query, the kind where a brand gets named or left out. Second, the Gemini link does not tell you which sources Siri uses. Search Engine Journal's analysis of the announcement makes the same point: Apple has not said how far Siri's answers track Gemini's, so visibility in one does not carry over to the other by default.
What does Apple say about source links?
Apple's only statement on source links sits in its Applebot documentation, published September 4, 2026. It says answers to broad world knowledge questions in Siri and Search "may include links to sources and websites used to help generate the answer." Neither newsroom post mentions links or citations, and MacRumors' Siri AI guide does not describe them either.
"May include" is all Apple commits to. It does not say when links appear, for which questions, or how many.
The same page gives you four controls over how Apple uses your content:
| Control | What it does | Effect on Siri answers |
|---|---|---|
Allow or disallow Applebot in robots.txt | Governs the crawler that powers search in Spotlight, Siri and Safari | Blocking it removes you from those search features and from the crawl that feeds answers |
Disallow Applebot-Extended | Opts your content out of training Apple's foundation models | None on search; Apple says this token does not crawl and is not used in Search ranking |
nosnippet robots meta tag or X-Robots-Tag | Stops Applebot generating a description or web answer from the page | Keeps the page out of world knowledge answers, while it can still be crawled and found |
isAccessibleForFree: false in JSON-LD | Marks paywalled content | The page stays eligible for search, but Apple does not use it as context for AI answers |
The table shows the trade-off Apple offers. You can refuse training without leaving search, and you can stay in search while keeping a page out of answers. The reverse, appearing in answers while blocking Applebot, is not on offer.
What is still unknown about how brands get cited?
Almost everything past the crawl. These questions stay open after reading Apple's Applebot page, both launch posts and the main coverage:
- How Siri selects sources: Apple's Applebot page lists search ranking factors (aggregated user engagement, relevance, the number and quality of inbound links, approximate location and page design), but it presents them for search results and says nothing about Siri's generated answers.
- When links appear: Apple says only "may include", with no examples or thresholds.
- Whether a Siri link sends a referrer: if a link opens in Safari without one, the visit lands in analytics as direct traffic, and Search Engine Journal found no Apple statement on this.
- Whether any reporting exists: Apple offers nothing comparable to Search Console for Siri answers.
- Which index Siri reads: Apple describes Applebot data powering Spotlight, Siri and Safari, and says Siri's answers draw on information from across the web. It does not say whether that means Apple's own index only.
Treat any vendor that claims a Siri ranking formula with suspicion until Apple documents one. What you can do now is make sure you are eligible and watch what the answers say.
How do you check that Applebot can reach your site?
Read your robots.txt the way Apple says Applebot reads it, then confirm real Applebot visits in your logs. Apple's rules differ from other crawlers in one place: if your file has no Applebot group but does have a Googlebot group, Applebot follows the Googlebot rules.
- Find the group Applebot obeys. Look for a
User-agent: Applebotgroup. If there is none, look forUser-agent: Googlebot; Applebot follows it. If neither exists, theUser-agent: *group applies. - Decide on Applebot-Extended on its own. Disallowing it opts out of model training and leaves search untouched. Make that call on its own merits, separately from any blanket AI-bot block.
- Check your CDN and firewall rules. Bot-management settings can block a crawler that robots.txt allows. Our Cloudflare AI crawler guide shows how one vendor's defaults affected Googlebot; read your own vendor's verified-bot list for Applebot.
- Verify real visits. Applebot's search requests carry a Safari-style user agent that, in Apple's current examples, ends in
(Applebot/0.1; +http://www.apple.com/go/applebot). Anyone can copy that string, so confirm the IP as well: real Applebot traffic resolves to a*.applebot.apple.comhostname on reverse DNS, and the forward lookup of that hostname returns the same IP. You can also match IPs against the CIDR ranges in Apple'sapplebot.jsonfile. - Leave rendering resources open. Apple says Applebot may render pages in a browser, and blocked JavaScript or CSS can leave the render incomplete.
- Audit
nosnippetand paywall markup. Make sure pages you want in answers do not carry either by accident, for example through a template-wide meta tag.
The free AI crawler access checker gives a verdict from your robots.txt for 44 AI crawlers, Applebot and Applebot-Extended included, with the rule that decided each one. It applies Apple's Googlebot fallback, so when your file has no Applebot group, the Applebot row shows the Googlebot rule it follows.
Access is the first stage of four. A crawl does not guarantee a citation, as crawl vs citation explains stage by stage, and the AI crawler guide covers the other bots worth allowing.
How do you test Siri's answers for your category?
By hand, with a fixed prompt list and a log. Citlyze does not track Siri today, and Apple offers no reporting, so a manual baseline is the honest option.
You need an Apple Intelligence device on iOS 27 or macOS 27, set to English. On iPhone and iPad that means a device outside the EU; a Mac works in the EU.
- Write 10 to 20 prompts a buyer would ask. Use the questions you already track elsewhere, such as "best payroll software for a 20-person restaurant" or "is [your brand] good for small teams". If you do not have a list yet, start from buyer prompts for AI search.
- Ask each prompt in a new conversation. Type it where your device allows, so speech recognition does not change the wording.
- Log the answer before you tap anything. Record which brands Siri names, in what order, and whether the answer shows source links.
- Open each source link. Note the domain, and check later whether the visit appears in your analytics and with what referrer.
- Repeat on three different days. AI answers vary between runs of the same question, so one pass is a sample of one.
- Ask the same prompts in ChatGPT and Gemini. Compare which brands and sources overlap with Siri, rather than assuming they match.
A simple log keeps the runs comparable:
| Prompt | Date and run | Brands named, in order | Links shown | Linked domains | Referrer in analytics |
|---|---|---|---|---|---|
| best payroll software for restaurants | 2026-10-12, run 1 | Brand A, Brand B, you | Yes | vendor.com, review-site.com | Not yet checked |
Keep the limits in mind. Apple caps daily use of server-side features, so a long session may stop partway. Three runs of fifteen prompts gives you a baseline, too few runs to show a trend. It tells you whether Siri names you at all and which sources it leans on, which is enough to decide what to fix.
Is Siri AI worth optimizing for yet?
It is worth an hour, not a project. The work that makes you eligible for Siri is the same work every answer engine rewards: open access for the search crawler, pages that state facts in plain words, and third-party sources that mention you. None of it is Siri-specific, so none of it is wasted if Siri stays small.
Two things would change that judgment. If Apple documents how sources are chosen, or adds reporting, Siri becomes measurable, and it is worth tracking like any other engine. If the EU and the October languages bring Siri AI to most iPhones, the audience becomes too large to leave to a quarterly manual check.
What to do this week
Open your robots.txt and find the group Applebot obeys, including the Googlebot fallback. Confirm Applebot in your server logs by reverse DNS. Then run ten buyer prompts through Siri and write down whose names come back. The free AI crawler access checker handles the first step in under a minute. Citlyze tracks the other engines, from ChatGPT to Google AI Mode, with a 14-day free trial that needs no card.