Update channels/source resolution

lib/plausible/imported/google_analytics4.ex

@@ -1,3 +1,3 @@
 defmodule Plausible.Imported.GoogleAnalytics4 do
   @moduledoc """
   Import implementation for Google Analytics 4.
@@ -352,6 +352,6 @@
   defp parse_source("duckduckgo"), do: "DuckDuckGo"
 
   defp parse_source(ref) do
-    Plausible.Ingestion.Source.parse("https://" <> ref)
+    Plausible.Ingestion.Source.from_referrer("https://" <> ref)
   end
 end

lib/plausible/ingestion/acquisition.ex

@@ -9,7 +9,8 @@ defmodule Plausible.Ingestion.Acquisition do
 
   Notable differences from GA4 that have been implemented just for Plausible:
   1. The @custom_source_categories module attribute contains a list of custom source categories that we have manually
-  added based on our own judgement and user feedback. For example we treat AI tools (ChatGPT, Perplexity) as search engines.
+  added based on our own judgement and user feedback. For example we group AI assistants (ChatGPT, Claude, Gemini, etc.)
+  into their own "AI Assistants" channel.
   2. Google is in a privileged position to analyze paid traffic from within their own network. The biggest use-case is auto-tagged adwords campaigns.
   We do our best by categorizing as paid search when source is Google and the url has `gclid` parameter. Same for source Bing and `msclkid` url parameter.
   3. The @paid_sources module attribute in Plausible.Ingestion.Source contains a list of utm_sources that we will automatically categorize as paid traffic
@@ -30,13 +31,23 @@ defmodule Plausible.Ingestion.Acquisition do
     {"vkontakte", "SOURCE_CATEGORY_SOCIAL"},
     {"threads", "SOURCE_CATEGORY_SOCIAL"},
     {"ecosia", "SOURCE_CATEGORY_SEARCH"},
-    {"perplexity", "SOURCE_CATEGORY_SEARCH"},
+    {"kagi", "SOURCE_CATEGORY_SEARCH"},
     {"brave", "SOURCE_CATEGORY_SEARCH"},
-    {"chatgpt.com", "SOURCE_CATEGORY_SEARCH"},
     {"temu.com", "SOURCE_CATEGORY_SHOPPING"},
     {"discord", "SOURCE_CATEGORY_SOCIAL"},
     {"sogou", "SOURCE_CATEGORY_SEARCH"},
-    {"microsoft teams", "SOURCE_CATEGORY_SOCIAL"}
+    {"microsoft teams", "SOURCE_CATEGORY_SOCIAL"},
+    {"bluesky", "SOURCE_CATEGORY_SOCIAL"},
+    {"mastodon", "SOURCE_CATEGORY_SOCIAL"},
+    {"chatgpt", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"claude", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"phind", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"deepseek", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"microsoft copilot", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"grok", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"google gemini", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"perplexity", "SOURCE_CATEGORY_AI_ASSISTANTS"},
+    {"x (twitter)", "SOURCE_CATEGORY_SOCIAL"}
   ]
   @source_categories Application.app_dir(:plausible, "priv/ga4-source-categories.csv")
                      |> File.read!()
@@ -69,6 +80,7 @@ defmodule Plausible.Ingestion.Acquisition do
       organic_shopping?(source, utm_campaign) -> "Organic Shopping"
       organic_social?(source, utm_medium) -> "Organic Social"
       organic_video?(source, utm_medium) -> "Organic Video"
+      ai_assistants?(source) -> "AI Assistants"
       search_source?(source) -> "Organic Search"
       email?(source, utm_source, utm_medium) -> "Email"
       affiliates?(utm_medium) -> "Affiliates"
@@ -188,6 +200,10 @@ defmodule Plausible.Ingestion.Acquisition do
     @source_categories[source] == "SOURCE_CATEGORY_EMAIL"
   end
 
+  defp ai_assistants?(source) do
+    @source_categories[source] == "SOURCE_CATEGORY_AI_ASSISTANTS"
+  end
+
   defp shopping_campaign?(utm_campaign) do
     Regex.match?(~r/^(.*(([^a-df-z]|^)shop|shopping).*)$/, utm_campaign)
   end

lib/plausible/ingestion/source.ex

@@ -15,6 +15,11 @@ defmodule Plausible.Ingestion.Source do
                 |> then(&["adwords" | &1])
                 |> MapSet.new()
 
+  @custom_source_suffixes %{
+    "officeapps.live.com" => "Microsoft 365",
+    "wikipedia.org" => "Wikipedia"
+  }
+
   @external_resource "priv/ref_inspector/referers.yml"
   @referers_yaml Application.app_dir(:plausible, "priv/ref_inspector/referers.yml")
 
@@ -36,11 +41,12 @@ defmodule Plausible.Ingestion.Source do
       |> Map.put(String.downcase(val), val)
     end)
 
+  @spec from_custom_sources(String.t()) :: String.t() | nil
   for {k, v} <- Enum.sort(lookup) do
-    def src(unquote(k)), do: unquote(v)
+    def from_custom_sources(unquote(k)), do: unquote(v)
   end
 
-  def src(_), do: nil
+  def from_custom_sources(_), do: nil
 
   def paid_sources() do
     @paid_sources |> MapSet.to_list()
@@ -64,39 +70,60 @@ defmodule Plausible.Ingestion.Source do
   perfectly, but at least we're making an effort for the most commonly used ones. For example, `ig -> Instagram` and `adwords -> Google`.
   """
   def resolve(request) do
-    tagged_source =
-      request.query_params["utm_source"] ||
-        request.query_params["source"] ||
-        request.query_params["ref"]
-
-    source =
-      cond do
-        tagged_source -> tagged_source
-        has_valid_referral?(request) -> parse(request.referrer)
-        true -> nil
-      end
-
-    find_mapping(source)
+    cond do
+      tagged = tagged_param(request) -> canonical(tagged)
+      has_valid_referral?(request) -> from_referrer(request.referrer)
+      true -> nil
+    end
   end
 
-  def parse(ref) do
-    case RefInspector.parse(ref).source do
-      :unknown ->
-        uri = URI.parse(String.trim(ref))
-        format_referrer_host(uri)
-
-      source ->
-        source
+  @doc """
+  Resolves a source from a bare referrer URL. This is also the entry point used
+  when importing referrers from Google Analytics, so that imported data 
+  is normalized identically to live ingestion.
+  """
+  def from_referrer(referrer) do
+    host =
+      referrer
+      |> String.trim()
+      |> URI.parse()
+      |> format_referrer_host()
+
+    downcased_host = String.downcase(host)
+
+    # Prefer custom source overrides over RefInspector so subdomain matches win
+    # (e.g. gemini.google.com resolves to Google Gemini, not Google). Exact host
+    # matches take priority, then suffix families, then RefInspector.
+    cond do
+      name = from_custom_sources(downcased_host) ->
+        name
+
+      name = from_custom_source_suffix(downcased_host) ->
+        name
+
+      true ->
+        case RefInspector.parse(referrer).source do
+          :unknown -> host
+          # Normalize RefInspector names through our aliases (e.g. Twitter to X (Twitter)).
+          name -> canonical(name)
+        end
     end
   end
 
-  def find_mapping(nil), do: nil
+  defp from_custom_source_suffix(host) do
+    Enum.find_value(@custom_source_suffixes, fn {suffix, name} ->
+      if host == suffix or String.ends_with?(host, "." <> suffix), do: name
+    end)
+  end
 
-  def find_mapping(source) do
-    case src(String.downcase(source)) do
-      name when is_binary(name) -> name
-      _ -> source
-    end
+  defp tagged_param(request) do
+    request.query_params["utm_source"] ||
+      request.query_params["source"] ||
+      request.query_params["ref"]
+  end
+
+  defp canonical(source) do
+    from_custom_sources(String.downcase(source)) || source
   end
 
   def format_referrer(request) do

lib/plausible_web/plugs/favicon.ex

@@ -34,12 +34,25 @@ defmodule PlausibleWeb.Favicon do
   @external_resource @placeholder_icon_location
   @custom_icons %{
     "Brave" => "search.brave.com",
+    "Kagi" => "kagi.com",
     "Sogou" => "sogou.com",
     "Wikipedia" => "en.wikipedia.org",
     "Discord" => "discord.com",
     "Perplexity" => "perplexity.ai",
     "Microsoft Teams" => "microsoft.com",
-    "LinkedIn" => "linkedin.com"
+    "LinkedIn" => "linkedin.com",
+    "Linktree" => "linktr.ee",
+    "Bluesky" => "bsky.app",
+    "Mastodon" => "mastodon.social",
+    "Google Gemini" => "gemini.google.com",
+    "ChatGPT" => "chatgpt.com",
+    "Claude" => "claude.ai",
+    "Phind" => "phind.com",
+    "DeepSeek" => "deepseek.com",
+    "Microsoft Copilot" => "copilot.microsoft.com",
+    "Grok" => "grok.com",
+    "X (Twitter)" => "x.com",
+    "Microsoft 365" => "office.com"
   }
 
   def init(_) do