Why AI Search Engines Treat Ratings as a Trust Signal
When someone asks ChatGPT or Perplexity to recommend a product, those systems are not simply picking the most popular result from Google. They are synthesising signals from across the web, and one of the clearest, most machine-readable signals you can give them is a structured rating. Not a paragraph that says "our customers love us" - an actual AggregateRating schema block that tells an AI, with numbers, exactly how well-regarded your product is.
AI search engines are built on large language models that were trained on structured and unstructured web data. Structured data - especially schema markup - gives these models something reliable to parse. A ratingValue of 4.8 with a reviewCount of 312 is unambiguous. It requires no inference. That kind of clarity is exactly what AI systems prefer when forming a recommendation.
This is not theoretical. Perplexity, for example, frequently surfaces product recommendations with explicit star ratings and review counts drawn directly from structured data on the source page. If your page lacks that markup, it is invisible to that process, regardless of how many genuine reviews you have sitting in plain text.
What AggregateRating Schema Actually Contains
The AggregateRating type is defined by Schema.org and represents a rolled-up rating based on multiple individual ratings or reviews. It sits as a property inside another schema type - most commonly Product, LocalBusiness, Course, or Recipe.
Here are the core properties you need to understand:
- ratingValue: The average rating, expressed as a number (e.g. 4.7). This can be a decimal.
- reviewCount: The total number of written reviews. This is distinct from the number of ratings.
- ratingCount: The total number of ratings, including those without a written review.
- bestRating: The highest value on your rating scale. Defaults to 5 if omitted, but you should always include it explicitly.
- worstRating: The lowest value. Defaults to 1 if omitted. Again, explicit is better.
- itemReviewed: A reference to the entity being rated. This links the rating back to the parent object.
A common mistake is including only ratingValue and nothing else. That gives AI systems very little to work with. A complete block with all of the above properties is far more informative and far more likely to be used.
How to Write the JSON-LD
JSON-LD is the recommended format for schema markup, and it is what AI crawlers tend to find most reliably. Here is a practical example for a product page:
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Merino Wool Base Layer",
"description": "Lightweight merino wool base layer for hiking and outdoor sports.",
"brand": {
"@type": "Brand",
"name": "Ridgeline Outdoors"
},
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": "4.8",
"reviewCount": "214",
"ratingCount": "289",
"bestRating": "5",
"worstRating": "1"
}
}Place this inside a <script type="application/ld+json"> tag in the <head> of your page or just before the closing </body> tag. Both locations work. The head is marginally preferred because crawlers hit it earlier.
A few things worth noting. The ratingValue should reflect the live average from your actual reviews. Do not hardcode a value and forget to update it. If your platform recalculates ratings dynamically, make sure your schema block does too. Static schema that drifts away from your real data is worse than no schema at all, because it erodes trust with both users and AI systems that cross-reference values.
Nesting AggregateRating Inside Other Schema Types
The power of AggregateRating comes from its relationship to the parent entity. On its own it does not mean much. Nested inside a Product block, it tells AI systems "this specific product has this specific rating." That specificity is what gets you cited.
Product pages
This is the most common use case and the highest-impact one for e-commerce brands. Every product page with more than, say, ten reviews should have a fully populated AggregateRating block. If you have 500 product pages and none of them have this markup, you are handing a significant AI visibility advantage to competitors who do.
LocalBusiness pages
If you have a physical location or service area, nesting AggregateRating inside a LocalBusiness schema block is enormously valuable. AI tools like ChatGPT are frequently asked "what is the best plumber in Bristol?" or "which florist near me has the best reviews?" If your LocalBusiness schema includes a solid aggregate rating, you become a viable answer to those queries.
Course and service pages
For training providers and service businesses, the same logic applies. A Course with a 4.9 rating from 87 reviews, expressed clearly in structured data, stands out against a competitor's page that has the same reviews buried in a comments section with no schema. AI systems will pull the structured signal almost every time.
For a deeper look at how course schema works in practice, see our post on how to use Course schema to get your training cited by AI search.
The Relationship Between Review Volume and AI Citation Likelihood
There is a practical threshold effect at play here. A ratingValue of 4.9 from 3 reviews does not carry the same weight as 4.7 from 340 reviews. AI systems, like humans, factor in sample size when assessing reliability. A higher reviewCount increases the credibility of the score and makes the recommendation less risky for the AI to surface.
This does not mean you should artificially inflate review counts. It means you should actively encourage legitimate reviews from real customers and make sure that as that number grows, your schema reflects it in real time. A review count that lags six months behind reality is a missed opportunity.
There is also a quality dimension. AI systems trained on web data have seen enough fake review patterns to be sceptical of 5.0 averages with suspiciously round review counts. A 4.6 with 180 reviews often reads as more credible than a 5.0 with 200. Do not chase perfection. Chase authenticity and volume.
Common Implementation Errors That Undermine Your Ratings
Schema markup mistakes are more common than most people realise, and several of them specifically hurt AggregateRating performance.
Missing bestRating and worstRating
If you omit these, Schema.org defaults to a 1-5 scale. But if your platform actually uses a 1-10 scale and you are reporting a ratingValue of 8.4, an AI that assumes a 5-point scale will interpret that as an off-the-charts rating. This kind of ambiguity can lead to your data being discarded. Always declare your scale explicitly.
Applying AggregateRating to a page, not an entity
The AggregateRating must be attached to a specific entity - a product, a business, a course. If you apply it loosely to a generic page with no clear parent entity, you lose the specificity that makes it useful to AI systems. The entity needs a name, ideally a description, and ideally a URL.
Stale data
As mentioned, static schema that does not update when your reviews change is a liability. If your Shopify or WooCommerce theme generates schema from a template, check whether that template pulls live review data or hardcoded values. Hardcoded values go stale fast.
For a broader look at issues that can quietly damage your AI visibility, our post on common schema markup mistakes that hurt AI visibility is worth reading alongside this one.
Duplicate schema blocks
Some platforms and plugins add schema automatically, which can result in two Product blocks on the same page - one from the theme, one from a plugin. If those blocks have conflicting ratingValue figures, crawlers and AI systems do not know which to trust. Audit your pages with Google's Rich Results Test or a tool like Schema Markup Validator to check for duplicates.
How Customer Reviews Amplify Your AggregateRating Signals
Schema markup and actual customer reviews are not separate strategies. They work together. The schema tells AI systems that the rating exists in a structured, machine-readable format. The underlying reviews, particularly if they appear on third-party platforms like Trustpilot, Google Reviews, or Yelp, provide corroboration that AI systems can cross-reference.
When an AI sees your AggregateRating schema showing 4.7 from 190 reviews, and it can also find references to your brand's ratings on independent review platforms, the confidence in that signal increases significantly. This is why review generation and schema implementation are best treated as a combined effort rather than isolated tasks.
We have written about this dynamic in more depth in our post on how customer reviews influence whether AI recommends your business.
Testing Your AggregateRating Implementation
Once your schema is live, test it before assuming it works. Use these tools:
- Google Rich Results Test (search.google.com/test/rich-results): Paste your URL or code snippet. It will flag errors and warnings in your
AggregateRatingblock specifically. - Schema Markup Validator (validator.schema.org): More thorough than Google's tool for Schema.org compliance. It catches issues that Google's tool sometimes misses.
- Bing Webmaster Tools: Often overlooked, but Bing's structured data testing gives you another perspective, and Bing's index feeds into some AI tools.
After testing, give it two to four weeks and then query AI search engines directly. Ask ChatGPT, Perplexity, or Gemini about your product category and see whether your brand or products surface in the recommendations. If they do not, revisit the quality and completeness of your schema, your review volume, and whether your brand is mentioned on authoritative third-party sources.
If you want a structured view of where your site currently stands, a free AI visibility audit from FlinnSchema is a good starting point. It surfaces schema gaps, missing signals, and specific fixes ranked by impact.
Frequently Asked Questions
Does AggregateRating schema directly affect Google star ratings in search results?
Yes, but with conditions. Google will display star ratings in organic search results for certain page types - products, recipes, local businesses - when AggregateRating schema is correctly implemented. However, Google has tightened its eligibility rules and may not show stars for every category. AI search visibility is a separate and additional benefit that comes from the same markup.
Can I use AggregateRating schema if my reviews are hosted on a third-party platform?
You can include AggregateRating schema on your own pages even if the underlying reviews live on Trustpilot, Google, or another platform - provided the data you include accurately reflects your real aggregate score. Do not fabricate or inflate figures. If your third-party rating is 4.3, that is the number to use. Some brands pull this data via API to keep their schema current automatically.
How many reviews do I need before AggregateRating schema becomes worth implementing?
There is no hard minimum, but fewer than five reviews makes the signal weak. Practically speaking, ten or more reviews gives you a reasonable sample size that AI systems and search engines will treat with some credibility. Google's own documentation suggests a minimum of one review for eligibility, but from an AI influence perspective, aim for at least 20 to 30 before expecting meaningful impact.
Will AggregateRating schema alone get my product recommended by ChatGPT?
Schema markup is one signal among many. It significantly improves your chances, but AI recommendations also factor in brand mentions across the web, the authority of sources linking to you, the quality of your page content, and whether your brand appears in the training data of the model. Think of AggregateRating schema as a necessary condition, not a sufficient one on its own. Pair it with strong review volume, third-party mentions, and well-structured page content for the best results.