AI Trust Audit Update: The 7 Most Important Changes and What Shop Operators Should Do Now
The AI Trust Audit measures how well AI systems like ChatGPT, Perplexity, or Apple Intelligence can read, understand, and categorize your website. With this update, we are adapting the catalog to the current AI landscape. It is based on practical tests with large shops and new insights into how AI systems select sources.
Important note: A high score in the audit is no guarantee that AI systems will mention your brand. It shows whether you have created the technical and content-related requirements to do so.
Changes in points at a glance
| Test | Previous | New |
|---|---|---|
| Perplexity (PerplexityBot) | 1 point | 3 points |
| Apple Intelligence (Applebot-Extended) | – | 2 points (new) |
| DeepSeek | – | 2 points (new) |
| Meta AI | 4 points | 2 points |
| Search field markup (SearchAction) | 4 points | 1 point |
| Cohere | 1 point | dropped |
| Share buttons | 0 points (informative) | dropped |
The increases and decreases balance out exactly (+6 and −6 points). The total achievable score remains the same.
1. New in the catalog: Apple Intelligence and DeepSeek
Using the token Applebot-Extended in your robots.txt, you determine whether Apple is allowed to use your content for Apple Intelligence. If it is not specifically regulated, the rules for the classic Applebot apply. The new test shows whether you are—often unintentionally—locking yourself out of Apple's AI functions.
DeepSeek is increasingly being integrated into search and research tools. So far, there is no officially documented crawler. We therefore check whether your robots.txt excludes DeepSeek explicitly or via a general block.
In return, Cohere, which plays practically no role in end-customer purchasing decisions, and the share buttons, which have no influence on how AI systems read your page, have been removed.
2. Perplexity now counts for more
Perplexity identifies its sources visibly and links directly to them. For shops and brands, access for the PerplexityBot is therefore particularly valuable. Meta AI and search field markup, on the other hand, count for less: so far, Meta AI has generated hardly any direct shop visits, and SearchAction plays almost no role in AI responses.
3. Firewalls and bot protection are recognized
Some bot protection systems allow human visitors through but block automated requests—including those from AI systems. The audit now recognizes such blocks more reliably. What our testing crawler cannot see, AI crawlers usually cannot see either. This is often the biggest unnoticed lever.
4. Reviews that AI can actually read
Many shops only display star ratings via a reloaded widget. They are visible to humans, but not to many AI crawlers. AI systems then tend to answer questions about trust and goodwill using external sources. If a machine-readable overall rating is missing in the source code, the test card now shows a concrete next step—directly under "What you can do".
5. Real product pages and clear company pages
For product data, prices, and product identifiers (GTIN/EAN), the audit now specifically selects a real product page and skips FAQ, help, legal, and magazine pages. Additionally, we recognize product data that modern shop systems provide as embedded data in the source code. Even for the "About Us" page, accessibility explanations or partner programs are no longer incorrectly classified as the company page.
6. Protection against Wikipedia confusion
If your brand name shares a Wikipedia entry with another brand, it can confuse AI systems. The audit now detects such confusion and recommends clear links via organizational data (sameAs), such as to Wikidata or your review profile.
7. Important content first—and a uniform point scale
An analysis of over 50,000 ChatGPT citations shows: A large part of the cited passages comes from the top third of a page. Structured content such as lists, tables, or fact boxes now count fully if they appear early on the page. This is a practical guideline, not a fixed rule.
Furthermore, all reports are calculated according to the same current weights. Older and new audits are therefore directly comparable. Older reports may change slightly without anything having changed on your website; the individual test results remain intact.
Your Checklist
- Check robots.txt: Are PerplexityBot, Applebot-Extended, and the other AI crawlers regulated as you want them to be?
- Coordinate bot protection: Let reputable AI crawlers and our testing crawler through your firewall.
- Make reviews machine-readable: Store your overall rating as structured data in the source code.
- Deliver complete product data: Price, currency, availability, and GTIN should be readable without JavaScript.
- Start a new audit: See how your website performs according to the new tests.