P1 · Article server HTML
Sites: Prajavani (PV, www.prajavani.net, repo tpml) and Deccan Herald (DH, www.deccanherald.com, repo tpml-dh) · Priority: 1 of 6 · Prepared: 18 Nov 2026 by the Product, SEO and Data team · Internal refs: #1, #2, #7 (hero image part), #9, #11, #12, #31 (measurement), #32 (fix 1), #41
Summary
Both sites send article paragraphs as empty containers in the first HTML response. The text exists only in a JSON data script and appears after JavaScript runs. Google indexes from the first HTML and renders JavaScript later, if at all. On PV this has left an estimated 76% of print and web-desk articles indexed without body text. Ship the paragraph fix (item 1) on both sites first. It is built for PV and verified on the DH training host. In the same release, ship the other article-template changes that make the first HTML complete: the "more from this section" link block (item 2), the "Also read" link fix (item 3), the hero image (item 4) and the premium-story body (item 5). Items 2 to 5 touch the same templates, so one release saves repeat QA. Items 6 and 7 are already fixed in the same branch and only need checking at release. Then measure (item 8).
Impact
What changes for Google: the article body, the contextual links to other stories and the hero image are in the first HTML. Google no longer depends on a JavaScript render to see them.
What changes for readers: nothing visible is intended. Ad positions must stay the same (see item 1, Verify). Ad volume rises for in-site navigation (item 6).
Numbers (each with its source):
- PV, 16 Nov 2026 site:-restricted Google search test (663 indexed stories, one body phrase each): body not indexed for 482 of 634 resolved stories (76%). Print copies never shown in Google: 295 of 382 (77%). Print copies with impressions: 71 of 94 (76%). Web-desk
textstories with impressions: 69 of 97 (71%). PV print and web-desk stories were about 91% of PV output, 7 Sep to 5 Nov 2026 (Quintype API). - DH, 16 Nov 2026 test: 8 of 8 never-shown editorial stories and 2 of 2 controls found by body phrase. Earlier sample estimate: about 0 to 7% without body. DH has the same code. The likely reason it does better is that Google renders DH more often (see "Not proven").
- PV before print-to-web (Jan 2026; 41 stories resolved): 41% body not indexed (17 of 41). Small sample (about ±15 points).
Not proven:
- That the fix will raise traffic. There is no clean measure yet. PV article-landing sessions (GA4, Google organic incl. Discover) fell 36% from Aug 2024 to Nov 2024 and recovered by Jul 2025. The fall began in Sep 2024, before the code change (8 Oct 2024), and a Google core update landed in Nov 2024. The change cannot be blamed alone.
- The mechanism (limited rendering for PV) is a hypothesis. Crawl stats cannot measure it.
- The share for
syndicatedwire stories (about 6% of output) and small templates was not tested. It is likely similar because the code is the same. - Link availability in Google's saved copy was checked on three PV articles only. Site-wide frequency is not measured.
Items
1. Article paragraphs missing from server HTML — Critical · PV and DH · Status: DH fix live 19 Nov 2026; PV fix built, not pushed
Issue: the server sends every plain-text paragraph as an empty container. The text sits only in the JSON data <script> and appears after JavaScript runs.
<div id="text-element-with-ad"><div></div></div>
Evidence:
- PV, live fetch, 16 Nov 2026 and earlier: 62 of 62 article pages with text blocks had empty containers (templates text, syndicated-print, syndicated, pv-exclusive). An earlier 64-URL sample gave 42 of 42. Share of body text in server HTML on 3 measured articles: 0%, 25% (big-fact blocks only), 0%.
- DH, live fetch: 31 of 31 sampled articles empty, same markup.
- Two other Quintype publishers (The Quint, Swarajya) serve paragraphs in server HTML. The cause is the custom TPML component, not the platform.
- Google's saved HTML (Search Console "View crawled page", 16 Nov 2026) showed empty paragraph containers for three PV articles, identical to raw HTML. A control article had a filled copy and its body was searchable.
- Git history (PV
tpmlupstream/master, checked 16 Nov 2026): at launch (d0db20fd, 28 Feb 2023) text elements went through Quintype'sStoryElement, which renders paragraphs on the server. The client-side split arrived 8 Oct 2024 (f9d0a0e3, "add new custom story element component to render ad after 4th para":getParagraphsArrregex plususeEffect).d45f010f(15 Oct 2024) swapped the regex forDOMParser. - DH:
quintype/tpml-dhmain0bc245a1(23 Oct 2026):story-elements/index.jsusesDOMParseringetParagraphsArr(line 134) and module-levelcountPara,hasRenderedMidAd1/2(lines 236 to 238).
Cause: story-templates/story-elements/index.js, component CustomStoryElement. To insert ads between paragraphs, the text is split with the browser-only DOMParser, and <p> tags are counted inside a useEffect. Neither runs on the server.
Fix:
- PV: branch
fix/ssr-story-paragraphs, commitbcd5d909(local, not pushed). It adds a server-safe paragraph splitter and a per-story ad plan computed during render. Checked by the author: 143 stories / 633 text blocks give the identical split and ad positions; local server HTML has 0 empty containers (production has 1 to 8 per page); ad and Taboola positions are the same after JavaScript; hydration keeps all paragraph nodes. - DH: port the PV fix. A fixed build is on the DH training host (see Verify); its commit reference is not recorded in our files.
- Do not reintroduce
DOMParseror any browser-only API in the render path.
Impact: removes the dependence on Google's render step for article text. For PV this targets the 76% indexed-without-body share above. For DH it protects a currently better position and puts the "Also read" card into server HTML (see item 3). Traffic effect is not proven.
Verify:
- Status as of 18 Nov 2026: the tech team is testing on DH first, PV to follow (agreed 16 Nov 2026). DH fix verified on the training host (
deccanherald-training-web.quintype.io) on 18 Nov 2026: 3 stories fetched with Googlebot UA before JavaScript. Training: 29 of 29 body paragraphs in server HTML (16/16, 8/8, 5/5). Live DH: 2 of 29. H1, title, canonical, hero image, byline, NewsArticle JSON-LD and AMP link unchanged. HTML size 1.30 MB (live 1.29 MB). Training is blocked in robots.txt (Disallow: /) and canonicals point to www, so it cannot leak into Google. - Side effect seen on training:
og:typeis nowarticle(live:website). That belongs to P5. - DH production, 19 Nov 2026 (fix commit
tpml-dh250ef126, "ssr fix (#2110)", 18 Nov 2026). Server-HTML check, Googlebot user agent, before JavaScript: every freshly rendered story had 0 empty paragraph containers. On the same 3 stories, body paragraphs in server HTML went from 1/13, 1/15 and 0/10 before the deploy to 15/15 and 10/10 for two of them. The third story was an old Cloudflare cache copy; fetched fresh, it was complete. In a sample of 12 recent stories, 3 were still old cached copies (all Cloudflare HITs). The cache rules (s-maxage=900,stale-while-revalidate=1000,stale-if-error=14400) should age these out. Re-check on 20 Nov 2026 and purge the article cache if any remain. Fresh pages had one first and one second mid-article ad slot and no duplicate ad slot IDs. Premium stories now render the preview instead of a spinner (item 5).- All templates: checked 2 stories per template among the latest 600 (text, syndicated, dh-exclusive, brandspot-client-story, listicle, photo, horoscope, news-briefs, video, visual-story). All had 0 empty containers, and free stories had all their paragraphs except the "Also read" text.
- Horoscope: the daily sign texts are now in the server HTML, 12 of 12 (0 of 12 on 12 Nov 2026). Still open: no
<h1>/<h2>in that HTML, andarticleBodyin the structured data is still the site URL (P4 item 8, P5). - Old stories: stories from March 2024 are also fixed, because pages are built on each request.
- Not tested: live-blog, podcast, cartoon, panchanga, embed and immersive (none were recent).
- Points for the DH code (
story-elements/index.jsat250ef126), not seen live:- The mid-ad checks
currentPara === 1(line 147) andcurrentPara === 4(line 182) have no "already shown" guard. A heading, list or blockquote right after paragraph 1 or 4 would repeat the ad block with the same slot ID. isWebView(line 131) is always false on the server. For?app=truethe client then renders different markup, a hydration mismatch.html-utils.jsshows an HTML comment or stray closing tag as text, and counts a multi-paragraph blockquote as one paragraph.[Taboola Debug]console logs remain (lines 250 and 268).
- The mid-ad checks
- New with this release (DH): the e-paper promo after paragraph 1 is now in server HTML. Its
<h2>ANYTIME, ANYWHERE.</h2>is the first<h2>on every article (seen on…-4223637, 19 Nov 2026). Fix: inatoms/new-body-banner/index.js:44, renderepaper-headingas a<div>or<p>, keeping the class. Story headings stay the article's only<h2>s. - Still open on DH after this release: the "Also read" card is not in server HTML, and its desktop link is still
/${linkedStorySlug}(item 3). The hero image is still behindWithLazy. - PV training host, 19 Nov 2026 (Googlebot user agent, before JavaScript):
- Articles: 26 of 26 body paragraphs on 3 print and web-desk stories (6/6, 11/11, 9/9).
- Daily horoscope: 12 of 12 sign texts on 2 stories, the same as live PV, so no regression.
- Same as live and still open: horoscope has no
<h1>/<h2>(P4 item 8); the/https://host/<story>path returns 200, not 301 (P2).
- Repeat on PV: fetch an article with a Googlebot UA and JavaScript off. Expect 0 empty
text-element-with-adcontainers. Compare mid-article ad and Taboola positions with JavaScript on. - After release: re-run the same 3-story check on the live host. Then Search Console URL Inspection "View crawled page → HTML" on 3 stories. Then request indexing for priority URLs (top traffic plus affected list).
2. No related-story links in the server HTML of articles — High (PV), Medium (DH) · PV and DH · Status: fix proposed
Issue: the automatic "latest", "most read" and "more from section" blocks are not in the server HTML. Google's saved article HTML has few contextual links to other stories.
Evidence:
- Code, PV
tpmlupstream/master70d7bebf, DHtpml-dhupstream/main47a7e85e(audit 16 Nov 2026): the PV text template has no "more news" block (story-templates/listcle/index.js:83-85, commented out).LatestNewsandSideBarContentfetch only aftermakeSecondaryApiCalls, which is set after a real 700 px scroll (layouts/header/useScrollBehavior.js:138-139) plus a 2.75 s timer (story-elements/latest-news/index.js:47-54).LatestNewsis also wrapped inWithLazy(story-sidebar/index.js:56). - DH has the same gate (
latest-news/index.jsabout lines 57 to 67; "What Others Are Reading"section-stories/index.js:22-31). DH sets the flag on mount only at 1025 px or wider, so the smartphone crawler falls back to the scroll gate. - Live PV test, 16 Nov 2026 (headless Chrome as Googlebot, 412×9000 viewport, 8 s, no scroll, premium article
…-4306823): the only links to other stories were 68 Taboola links (16 distinct stories) and 17 footer links. No related, latest or most-read links. - Search Console crawled HTML, 16 Nov 2026, three PV articles: (a) Muhurat Trading: 17 article anchors to 13 stories, all in the footer; 16 of them in
hide-mobilegroups. (b) Prince William: one embedded "Also read" card with correct links, 17 footer anchors, no automatic block. (c) AI supply-chain: two valid "Also read" links plus 17 footer anchors. Taboola containers had no anchors in all three. - Search Console crawl stats, JSON file-type samples (PV export of 16 Nov 2026): Googlebot fetched
/api/v1/collections/most-readin 115 of 257 sampled JSON fetches. So Google sometimes triggers the widget calls, but the saved copies lack the blocks.
Fix: server-render a "more from this section" block of 5 to 10 links from the story's section collection, fetched in the article page loader, on every article, both sites. Use real <a href> links with the story headline. Keep the scroll-gated widgets for users.
Impact: gives each article a crawlable path to other stories. The effect on discovery speed and traffic is not measured. Low crawl per new URL (see P3) and never-shown articles (item 8) are context, not proof of cause.
Verify: view-source of an article shows at least 5 same-section story links. URL Inspection "View tested page" shows the block. The links are in the HTML before JavaScript and not only inside hide-mobile groups.
3. "Also read" card: doubled-host link and nested anchor — Medium (PV), Low (DH) · PV and DH · Status: fix proposed
Issue: the desktop "Also read" (ಇದನ್ನೂ ಓದಿ) card links to /https://www.prajavani.net/<story path>. That path returns 200 with the full article and a canonical to the real URL, so Google crawls and indexes a second copy.
Evidence:
- Googlebot HTML of
…/nitesh-tiwari-ramayana-odyssey-similarities-explanation-4383444(18 Nov 2026) containshref="/https://www.prajavani.net/entertainment/cinema/yash-ramayana-ravana-cinema-history-4165600". - Search Console Page indexing exports (data to 13 Nov 2026, 1,000-URL samples): doubled-host URLs are 81 of "Crawled – currently not indexed", 15 of "Alternate page with proper canonical", 90 of "Page with redirect". Story numbers reach 4392920 (Nov 2026), so the bug is current.
- URL Inspection, 18 Nov 2026: two doubled URLs (
…-4383444,…-4385118) are "Submitted and indexed" with Google's canonical = the doubled URL itself. The real articles are also indexed, so Google holds two copies. - Search Console, 19 Oct to 15 Nov 2026: 47 doubled-host URLs with impressions, 2,572 impressions, 70 clicks (page filter
contains /https:). DH: 0. - DH has the same line. Checked 18 Nov 2026: 55 of the 60 latest DH stories have an "Also read" element, but the card is not in DH server HTML (because of item 1). After JavaScript runs, the DH card is
<a href="/https://www.deccanherald.com/…">. The doubled path returns 301 to the real story, so it costs one redirect hop. - Two smaller defects, seen in a browser DOM copy on 18 Nov 2026: (1) the desktop card wraps
@quintype/arrowHeadline, which renders its own<a>, so there is an<a>inside an<a>(invalid HTML; not checked against raw server HTML). (2) The card imagesrchasw=undefined(cause inCardimagenot traced);srcsetis correct.
Cause: PV tpml app/isomorphic/components/story-templates/story-elements/also-read/index.js:51 (master 0b3f7227; line from ea2d8e98, 16 Oct 2023): <Link href={/${linkedStorySlug}}>. linkedStorySlug comes from getLinkedStory() (app/api/utils.js:65) and holds the full url, not a slug. The mobile card (same file, line 14) and the AMP card (app/server/amp/override/components/rows/story-elements/also-read/index.js:36) use the value as is and are correct. The desktop card is hide-mobile, but it is still in the HTML Googlebot reads. DH: tpml-dh ee69e96a, also-read/index.js:50.
Fix:
href={linkedStorySlug}on line 51 and rename the variable tolinkedStoryUrl. Same one-line change on DH.- Remove the nested anchor: pass a prop to
Headlinethat turns off its link, or render the headline as plain text inside the card link (as the mobile card does with a<span>). - Give
Cardimagea default width sow=undefineddoes not appear. - Ship the DH change in the same release as DH's item 1, because that release moves the card into server HTML.
- Doubled URLs that already exist: 301 them to the real story. That belongs to P2 One URL per story.
Impact: stops new doubled URLs on PV. For DH it removes a redirect hop on each link. The effect on rankings is not measured.
Verify: a story with one "Also read" element has no /https: in its server HTML and no <a> inside <a>. About 4 to 6 weeks after release (with P2's 301), Search Console pages filtered by contains /https: trend to 0 and the doubled URLs in Page indexing move to "Page with redirect" or drop out.
4. Hero image not in server HTML — Medium (PV) · PV · Status: fix proposed
Issue: the article hero image is client-only, so Google's first HTML has no hero <picture>.
Evidence:
- Code,
tpmlupstream/master70d7bebf(16 Nov 2026):story-templates/story-header/hero-image/index.js:14wraps the hero in<WithLazy margin="20px">. It renders nothing on the server. The image appears only after JavaScript runs and an IntersectionObserver fires. Nofetchpriorityis set anywhere. - Chrome DevTools trace of a mobile PV article, 16 Nov 2026: the LCP element is a Google ad image at 17.0 s (lab), because the article's own hero is not in server HTML. Real users: p75 LCP 3.6 s, CLS 0.33.
Fix: render the hero <img> directly on article pages, with no WithLazy above the fold. Add fetchpriority="high" and explicit width and height. The header layout shift (useDevice returns null until mount; min-height 456 to 184 px) belongs to P6 Page speed, ads and analytics.
Impact: gives Google the hero image in the first HTML and is the main LCP fix on mobile. The size of the LCP gain is not measured.
Verify: view-source shows the hero <img> with fetchpriority="high" and width/height. On training or staging, the mobile LCP element is the hero image, not an ad.
5. Premium stories: spinner instead of body — Low (PV), DH noted · PV and DH · Status: investigated, optional fix
Issue: for access === "subscription" (non ad-free), story-elements/index.js (about lines 887 to 894) shows <SpinnerLoader /> instead of the body until a 500 ms timer and client-side premium-card computation finish.
Evidence:
- PV, rendered as Googlebot (headless Chrome), 16 Nov 2026: premium story
/technology/tech-tips/paper-battery-…-4306823shows 115 of 397 words plus the subscribe prompt. The server truncates to 115 words for everyone (app/server/data-loaders/story-page-data.js:92@70d7bebf). Google and users get the same content, so there is no cloaking risk. - PV premium is 658 of 38,193 recent stories (1.7%), and 93% of them get Discover impressions (traffic funnel report, stories 7 Sep to 5 Nov 2026, Search Console to 13 Nov 2026).
- DH, 17 Nov 2026: 6 sampled premium stories had 0 to 13 words inside
<p>in server HTML (Googlebot UA). Their AMP copies carry 534 to 1,178 words with no paywall. That is an AMP matter, see P3 AMP.
Fix: after item 1 ships, confirm the server HTML of premium stories holds the truncated body (the preview) and not the spinner. Optional: restore hasPart paywall markup (deleted in seo-hard-gate.js:274,291; see P5 Structured data and head tags).
Impact: low, because premium is a small share. Not a priority within this package.
Verify: fetch a premium story with a Googlebot UA: preview text and subscribe prompt present, no spinner-only body.
6. Mid-article ad counters never reset — Low · PV and DH · Status: fixed in the PV #1 branch; check at release
Issue: countPara and hasRenderedMidAd1/2 were module-level and never reset, so mid-article ads showed only on the first story viewed in a browsing session.
Fix: done in fix/ssr-story-paragraphs (per-story plan; ads on the first story of each page). DH has the same module-level variables (lines 236 to 238 on 0bc245a1); the port in item 1 covers them.
Verify: navigate through 3 stories in one session. Mid-article ads show on each. Ad volume will rise for in-site navigation, so tell ad operations before release.
7. isWebView / ?app=true check inverted — Low · PV · Status: fixed in the PV #1 branch; check at release
Issue: isWebView was set to true when the URL did not contain app=true, and the second mid-article ad was hidden when true. It worked only because of a stale value.
Fix: removed in the #1 branch with behaviour unchanged: the app still shows the second mid ad. Product to decide the intended app behaviour.
Verify: open an article with and without ?app=true. Both show the second mid-article ad, as before.
8. Before and after measurement plan — Analysis
Issue: PV print copies (published by the Varta Setu print-to-web system, syndicated-print template, live from late Mar 2026) are 79% of PV output and 96% of the articles Google never shows. The body is missing in Google's copy at about the same rate whether stories are shown or not (item 1). The likely causes of low traction are the missing body plus low demand for hyperlocal items.
Evidence:
- Stories published 7 Sep to 5 Nov 2026 (Quintype API) with Search Console data to 13 Nov 2026: 30,177 of 38,193 PV stories (79%) are print copies; 63.7% had any web impression, 8.5 web clicks per story. PV
text: 4,749 stories, 95.6%, 61.9 clicks per story. 11,454 PV stories (30%) had no web impression after 7 days. DH, all templates: 95.4% with an impression, 103.4 clicks per story. - URL Inspection sample, 16 Nov 2026: 300 random PV stories, all 300 "Submitted and indexed". Indexing is not the gap.
- Search Console crawled HTML, 16 Nov 2026: body phrase missing for 9 of 10 never-shown PV print copies (small sample).
Plan:
- Ship items 1 to 5. Then compare the print-copy "any impression" share and clicks per story for the 4 weeks before and after release. Collect data with a window that starts after the release.
- About 4 weeks after release, re-run the body-phrase search test on the same sampled stories (about 200 paid search queries are budgeted). Compare with 76% body not indexed.
- Optional separating test: about 50 stories from early Mar 2026 against about 50 from mid-Apr 2026 (same age, before and after print-to-web), about 100 queries.
Hold until the comparison is done: no other change to print copies in the same window (removal, noindex, length rules, round-ups, news-sitemap rules, or spreading the 06:00–09:00 publishing burst), so the effect of the fix can be measured on its own.
Not recommended now: noindex or removal of print copies. They are unique content and the fix has not shipped. If print copies stay low after the fix, the next step is an editorial decision (minimum length, district round-ups), outside this package.
Release checklist
- PV: push and review branch
fix/ssr-story-paragraphs(bcd5d909). Add the changes for items 2 to 5 (same templates). - DH: port items 1 and 3, add item 2. Re-run the training-host check (29 of 29 on the 3 reference stories was the pass on 18 Nov 2026).
- QA on training or staging, Googlebot UA, JavaScript off: no empty
text-element-with-adcontainers; at least 5 same-section links; no/https:links; hero<img>present; premium stories show the preview and not the spinner. - QA with JavaScript on: same ad and Taboola positions as before; no hydration warnings; mid-article ads show on each story in a session; second mid ad shows with
?app=true. - Check page weight: DH training HTML was 1.30 MB against 1.29 MB live. Flag any large rise.
- Release PV and DH.
- Same day: re-run the 3-story server-HTML check on the live hosts; URL Inspection "View crawled page" on 3 stories per site; request indexing for priority URLs.
- After 4 weeks: body-phrase re-test and before/after traffic comparison (item 8).
Dependencies and related packages
- P2 One URL per story: 301 for any non-canonical story path, which cleans up existing doubled-host URLs (item 3). Item 3's code fix and P2's redirect should go out close together.
- P3 AMP: DH premium AMP copies carry full text while canonical HTML has almost none (item 5).
- P4 Listing pages and internal links: pagination, home-page lazy collections and breadcrumbs. Item 2 is the article-side half of the same internal-linking problem.
- P5 Structured data and head tags:
og:typeon DH, premiumhasPartmarkup. - P6 Page speed, ads and analytics: hero layout shift (
useDevice,min-heighttransition), ad-related LCP, and the ad volume change from item 6.