parse ai citations

Author: kjellberg

lib/serp_parser/google/config.rb

@@ -172,6 +172,24 @@ def self.config_block
             end
           end
 
+          component :ai_citation do
+            variant "standard", meta: { first_seen: "2026-12-18" } do
+              match ".Y3BBE a.H23r4e"
+              model SerpParser::Models::AiCitation
+              text :title, ".", processors: [ :text, :clean_text ]
+              url :url, attribute: "href"
+            end
+          end
+
+          element :ai_overview do
+            variant "standard", meta: { first_seen: "2026-12-18" } do
+              container '.FkX2oe[data-subtree="aimc"]'
+              required_children [ ".Y3BBE" ]
+              text :answer, ".Y3BBE", processors: [ :strip_inner_style, :clean_text ]
+              has_many :citations, component: :ai_citation
+            end
+          end
+
           element :related_searches do
             variant "filter_pills", meta: { first_seen: "2025-12-23" } do
               container "div.fBctee"

lib/serp_parser/google/parsers/element.rb

@@ -100,7 +100,7 @@ def extract_field_value(field_def)
             value = element.text
             element_ref = element
             processors.each do |p|
-              if p == :remove_span_elements || p == :find_description_node
+              if p == :remove_span_elements || p == :find_description_node || p == :strip_inner_style
                 element_ref = apply_processor(p, value, element_ref)
                 value = element_ref.text if element_ref.respond_to?(:text)
               else
@@ -135,6 +135,8 @@ def apply_processor(processor_name, value, element = nil)
             Processors.parse_date(value)
           when :remove_span_elements
             Processors.remove_span_elements(element) || element
+          when :strip_inner_style
+            Processors.strip_inner_style(element) || element
           when :extract_number
             Processors.extract_number(value)
           when :normalize_number
@@ -252,6 +254,9 @@ def extract_component_data(element, variant)
             # For FAQ citations, need a non-internal URL
             elsif model_class == SerpParser::Models::FaqCitation
               return nil unless data[:url] && !data[:url].to_s.start_with?("/")
+            # For AI citations, need a non-internal URL
+            elsif model_class == SerpParser::Models::AiCitation
+              return nil unless data[:url] && !data[:url].to_s.start_with?("/")
             end
 
             model_class.new(**data)

lib/serp_parser/google/search.rb

@@ -11,6 +11,12 @@ def initialize(html_or_doc, schema: nil)
         @registry = Config.registry
       end
 
+      def ai_overview
+        results = Parsers::Element.find_all(@doc, :ai_overview, @registry)
+        return nil if results.empty?
+        SerpParser::Models::AiOverviewResult.new(**results.first)
+      end
+
       def organic_results
         results = Parsers::Element.find_all(@doc, :organic_result, @registry)
         models = results.map do |data|
@@ -59,6 +65,7 @@ def search_information
 
       def to_h
         {
+          ai_overview: ai_overview&.to_h,
           sponsored_results: sponsored_results.map(&:to_h),
           organic_results: organic_results.map(&:to_h),
           faq_results: faq_results.map(&:to_h),

lib/serp_parser/models/ai_citation.rb

@@ -0,0 +1,33 @@
+require "uri"
+
+module SerpParser
+  module Models
+    class AiCitation
+      attr_reader :title, :url
+      attr_accessor :position
+
+      def initialize(**args)
+        @title = args[:title]
+        @url = args[:url]
+      end
+
+      def domain
+        return nil unless url
+        host = URI.parse(url).host
+        return nil unless host
+        host.gsub(/www\./, "")
+      rescue
+        nil
+      end
+
+      def to_h
+        {
+          "position" => @position,
+          "title" => title,
+          "domain" => domain,
+          "url" => url
+        }
+      end
+    end
+  end
+end

lib/serp_parser/models/ai_overview_result.rb

@@ -0,0 +1,20 @@
+module SerpParser
+  module Models
+    class AiOverviewResult
+      attr_reader :answer, :citations
+
+      def initialize(**args)
+        @answer = args[:answer]
+        @citations = args[:citations].is_a?(Array) ? SerpParser::Collection.new(args[:citations]) : args[:citations]
+      end
+
+      def to_h
+        ans = answer&.strip
+        {
+          "answer" => (ans && !ans.empty?) ? ans : nil,
+          "citations" => citations ? citations.map(&:to_h) : []
+        }
+      end
+    end
+  end
+end

lib/serp_parser/processors.rb

@@ -34,6 +34,18 @@ def self.normalize_number(string, decimals: false)
       end
     end
 
+    # Returns a copy of the node with <style> and <script> children removed,
+    # so that .text on the result excludes embedded CSS/JS content.
+    # @param element [Nokogiri::XML::Element]
+    # @return [Nokogiri::XML::Element]
+    def self.strip_inner_style(element)
+      return element unless element.respond_to?(:dup)
+      clone = Nokogiri::HTML.fragment(element.to_html)
+      clone.css("style, script").each(&:remove)
+      clone.css('[style*="display:none"]').each(&:remove)
+      clone
+    end
+
     # Remove span elements from description
     # @param element [Nokogiri::XML::Element]
     # @return [Nokogiri::XML::Element]

spec/files/google/2025-12-23-mobile-matkasse.json

@@ -1,4 +1,5 @@
 {
+  "ai_overview": null,
   "sponsored_results": [],
   "organic_results": [
     {

spec/files/google/2025-12-23-mobile-middagsfrid-rabattkod.json

@@ -1,4 +1,5 @@
 {
+  "ai_overview": null,
   "sponsored_results": [],
   "organic_results": [
     {

spec/files/google/2026-01-01-mobile-swedbank.json

@@ -1,4 +1,5 @@
 {
+  "ai_overview": null,
   "sponsored_results": [],
   "organic_results": [
     {

spec/files/google/2026-01-29-mobile-matkasse.json

@@ -1,4 +1,5 @@
 {
+  "ai_overview": null,
   "sponsored_results": [],
   "organic_results": [
     {