diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 2c7378b..b622941 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -112,6 +112,37 @@ jobs:
           python examples/python_fetchkit.py
         timeout-minutes: 2
 
+  live-tests:
+    name: Live Fetcher Tests
+    if: github.event_name == 'pull_request'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+      - uses: dtolnay/rust-toolchain@stable
+      - uses: Swatinem/rust-cache@v2
+      - name: Detect changed fetchers
+        id: detect
+        run: |
+          FILTERS=$(bash scripts/changed-fetcher-tests.sh "origin/${{ github.base_ref }}")
+          echo "filters=$FILTERS" >> "$GITHUB_OUTPUT"
+          if [ -z "$FILTERS" ]; then
+            echo "skip=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "skip=false" >> "$GITHUB_OUTPUT"
+          fi
+      - name: Run live tests
+        if: steps.detect.outputs.skip == 'false'
+        run: |
+          for filter in ${{ steps.detect.outputs.filters }}; do
+            echo "::group::Running $filter"
+            cargo test --features live-tests -p fetchkit "$filter" -- --nocapture
+            echo "::endgroup::"
+          done
+        timeout-minutes: 5
+        continue-on-error: true
+
   # Gate job for branch protection.
   check:
     name: Check
diff --git a/crates/fetchkit/Cargo.toml b/crates/fetchkit/Cargo.toml
index 397d7f3..9231802 100644
--- a/crates/fetchkit/Cargo.toml
+++ b/crates/fetchkit/Cargo.toml
@@ -13,6 +13,7 @@ readme = "../../README.md"
 [features]
 default = []
 bot-auth = ["dep:ed25519-dalek", "dep:base64", "dep:sha2", "dep:rand"]
+live-tests = []
 
 [dependencies]
 tokio = { workspace = true }
diff --git a/crates/fetchkit/tests/fetcher_live.rs b/crates/fetchkit/tests/fetcher_live.rs
new file mode 100644
index 0000000..dfd4626
--- /dev/null
+++ b/crates/fetchkit/tests/fetcher_live.rs
@@ -0,0 +1,373 @@
+//! Live integration tests for fetchers against real endpoints.
+//!
+//! Gated behind `--features live-tests` so they never run during normal `cargo test`.
+//! Each test module maps 1:1 to a fetcher source file; CI runs only the modules
+//! whose fetcher changed.
+//!
+//! Assertions are structural (field presence, non-empty content, expected substrings)
+//! rather than exact-match, so tests tolerate minor upstream changes.
+//!
+//! Network errors (DNS, timeout, blocked) are treated as skips, not failures —
+//! live tests should only fail on unexpected response structure, not infra issues.
+
+#![cfg(feature = "live-tests")]
+
+use fetchkit::{FetchError, FetchOptions, FetchRequest, FetchResponse, FetcherRegistry};
+
+/// Shared options for live tests — default everything, both conversions on.
+fn live_options() -> FetchOptions {
+    FetchOptions {
+        enable_markdown: true,
+        enable_text: true,
+        ..Default::default()
+    }
+}
+
+fn registry() -> FetcherRegistry {
+    FetcherRegistry::with_defaults()
+}
+
+/// Network errors that indicate infra problems, not fetcher bugs.
+fn is_network_error(err: &FetchError) -> bool {
+    matches!(
+        err,
+        FetchError::FirstByteTimeout
+            | FetchError::BlockedUrl
+            | FetchError::ConnectError(_)
+            | FetchError::ClientBuildError(_)
+            | FetchError::RequestError(_)
+    )
+}
+
+/// Fetch and return Ok(response), or skip the test if the error is network-related.
+async fn fetch_or_skip(url: &str) -> Option<FetchResponse> {
+    let req = FetchRequest::new(url);
+    match registry().fetch(req, live_options()).await {
+        Ok(resp) => Some(resp),
+        Err(e) if is_network_error(&e) => {
+            eprintln!("SKIPPED (network): {url} — {e}");
+            None
+        }
+        Err(e) => panic!("unexpected fetcher error for {url}: {e}"),
+    }
+}
+
+/// Like fetch_or_skip but with as_markdown set.
+async fn fetch_markdown_or_skip(url: &str) -> Option<FetchResponse> {
+    let req = FetchRequest::new(url).as_markdown();
+    match registry().fetch(req, live_options()).await {
+        Ok(resp) => Some(resp),
+        Err(e) if is_network_error(&e) => {
+            eprintln!("SKIPPED (network): {url} — {e}");
+            None
+        }
+        Err(e) => panic!("unexpected fetcher error for {url}: {e}"),
+    }
+}
+
+// ---------------------------------------------------------------------------
+// github_repo
+// ---------------------------------------------------------------------------
+mod live_github_repo {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_repo_metadata() {
+        let Some(resp) = fetch_or_skip("https://github.com/rust-lang/rust").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.contains("rust-lang/rust") || content.to_lowercase().contains("rust"),
+            "content should mention the repo"
+        );
+        assert!(!content.is_empty());
+    }
+}
+
+// ---------------------------------------------------------------------------
+// github_issue
+// ---------------------------------------------------------------------------
+mod live_github_issue {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_issue() {
+        // Well-known issue: rust-lang/rust#1 (the very first issue)
+        let Some(resp) = fetch_or_skip("https://github.com/rust-lang/rust/issues/1").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(!content.is_empty());
+    }
+}
+
+// ---------------------------------------------------------------------------
+// github_code
+// ---------------------------------------------------------------------------
+mod live_github_code {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_source_file() {
+        let Some(resp) =
+            fetch_or_skip("https://github.com/rust-lang/rust/blob/master/README.md").await
+        else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("rust"),
+            "README should mention Rust"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// twitter
+// ---------------------------------------------------------------------------
+mod live_twitter {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_tweet() {
+        // Rust lang announcement tweet — stable, public
+        let Some(resp) = fetch_or_skip("https://x.com/rustlang/status/1821986021505405014").await
+        else {
+            return;
+        };
+
+        // Twitter APIs are unreliable; accept any non-panic response as proof
+        // the fetcher handled it. Only assert structure on 200.
+        if resp.status_code == 200 {
+            assert!(resp.content.is_some());
+        }
+    }
+}
+
+// ---------------------------------------------------------------------------
+// stackoverflow
+// ---------------------------------------------------------------------------
+mod live_stackoverflow {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_question() {
+        // "What is a NullPointerException" — one of the most famous SO questions
+        let Some(resp) = fetch_or_skip(
+            "https://stackoverflow.com/questions/218384/what-is-a-nullpointerexception-and-how-do-i-fix-it",
+        ).await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("null"),
+            "content should mention null"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// package_registry
+// ---------------------------------------------------------------------------
+mod live_package_registry {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_crate() {
+        let Some(resp) = fetch_or_skip("https://crates.io/crates/serde").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("serde"),
+            "content should mention serde"
+        );
+    }
+
+    #[tokio::test]
+    async fn fetches_pypi_package() {
+        let Some(resp) = fetch_or_skip("https://pypi.org/project/requests/").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("requests"),
+            "content should mention requests"
+        );
+    }
+
+    #[tokio::test]
+    async fn fetches_npm_package() {
+        let Some(resp) = fetch_or_skip("https://www.npmjs.com/package/express").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("express"),
+            "content should mention express"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// wikipedia
+// ---------------------------------------------------------------------------
+mod live_wikipedia {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_article() {
+        let Some(resp) =
+            fetch_or_skip("https://en.wikipedia.org/wiki/Rust_(programming_language)").await
+        else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("rust"),
+            "article should mention Rust"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// youtube
+// ---------------------------------------------------------------------------
+mod live_youtube {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_video_metadata() {
+        // "Me at the zoo" — first YouTube video ever, very stable
+        let Some(resp) = fetch_or_skip("https://www.youtube.com/watch?v=jNQXAC9IVRw").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(!content.is_empty());
+    }
+}
+
+// ---------------------------------------------------------------------------
+// arxiv
+// ---------------------------------------------------------------------------
+mod live_arxiv {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_paper() {
+        // "Attention Is All You Need"
+        let Some(resp) = fetch_or_skip("https://arxiv.org/abs/1706.03762").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("attention"),
+            "paper should mention attention"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// hackernews
+// ---------------------------------------------------------------------------
+mod live_hackernews {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_story() {
+        // HN item 1 — the very first story
+        let Some(resp) = fetch_or_skip("https://news.ycombinator.com/item?id=1").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(!content.is_empty());
+    }
+}
+
+// ---------------------------------------------------------------------------
+// rss_feed
+// ---------------------------------------------------------------------------
+mod live_rss_feed {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_rss() {
+        // Rust blog RSS feed
+        let Some(resp) = fetch_or_skip("https://blog.rust-lang.org/feed.xml").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("rust"),
+            "Rust blog feed should mention Rust"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// docs_site
+// ---------------------------------------------------------------------------
+mod live_docs_site {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_docs_rs() {
+        let Some(resp) = fetch_or_skip("https://docs.rs/serde/latest/serde/").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.to_lowercase().contains("serde"),
+            "docs.rs page should mention serde"
+        );
+    }
+}
+
+// ---------------------------------------------------------------------------
+// default (generic HTTP)
+// ---------------------------------------------------------------------------
+mod live_default {
+    use super::*;
+
+    #[tokio::test]
+    async fn fetches_plain_html() {
+        let Some(resp) = fetch_markdown_or_skip("https://example.com").await else {
+            return;
+        };
+
+        assert_eq!(resp.status_code, 200);
+        let content = resp.content.expect("should have content");
+        assert!(
+            content.contains("Example Domain"),
+            "example.com should contain 'Example Domain'"
+        );
+        assert_eq!(resp.format, Some("markdown".to_string()));
+    }
+}
diff --git a/scripts/changed-fetcher-tests.sh b/scripts/changed-fetcher-tests.sh
new file mode 100755
index 0000000..56de3ca
--- /dev/null
+++ b/scripts/changed-fetcher-tests.sh
@@ -0,0 +1,43 @@
+#!/usr/bin/env bash
+# Determine which live_* test modules to run based on changed fetcher files.
+#
+# Usage: scripts/changed-fetcher-tests.sh <base-ref>
+# Output: space-separated test name filters, or empty string if none changed.
+#
+# Maps: crates/fetchkit/src/fetchers/<name>.rs → live_<name>
+# Also maps: tests/fetcher_live.rs → (all)
+
+set -euo pipefail
+
+BASE_REF="${1:-origin/main}"
+
+# Get changed files relative to base
+CHANGED=$(git diff --name-only "$BASE_REF"...HEAD 2>/dev/null || git diff --name-only "$BASE_REF" HEAD)
+
+FILTERS=()
+RUN_ALL=false
+
+while IFS= read -r file; do
+  # If the live test file itself changed, run everything
+  if [[ "$file" == "crates/fetchkit/tests/fetcher_live.rs" ]]; then
+    RUN_ALL=true
+    break
+  fi
+
+  # Match fetcher source files
+  if [[ "$file" =~ ^crates/fetchkit/src/fetchers/([a-z_]+)\.rs$ ]]; then
+    name="${BASH_REMATCH[1]}"
+    # Skip mod.rs — user said no need to run all for shared changes
+    [[ "$name" == "mod" ]] && continue
+    FILTERS+=("live_${name}")
+  fi
+done <<< "$CHANGED"
+
+if $RUN_ALL; then
+  echo "live_"
+elif [[ ${#FILTERS[@]} -gt 0 ]]; then
+  # Deduplicate
+  printf '%s\n' "${FILTERS[@]}" | sort -u | tr '\n' ' '
+else
+  echo ""
+fi