26. PV AMP copies take ~46% of successful crawl — keep or retire AMP — High (PV), decision (raised 16 Nov 2026)
⚠️ Corrected below (AMP traffic, 16 Nov 2026): page-level /amp/ counts under-state AMP traffic, because Search Console usually credits AMP clicks to the canonical URL. Use the search-appearance figures. Step 1 done (16 Nov 2026): for stories published 7 Sep – 5 Nov 2026, AMP URLs earned 18,011 of 896,806 web clicks (2.0%), 8,692 of 18.6M Discover clicks (0.05%) and 153 Google News clicks (Search Console page-level, 7 Sep – 13 Nov 2026; checks/traffic_funnel/report_pv_2026-11-13.md). AMP costs ~46% of successful crawl for ~2% of new-article search clicks: retire it after #1 has been live ~4 weeks.
Added 12 Nov 2026. Not a bug: a decision to make after #1 ships.
Current AMP traffic (last 28 days, 20 Oct – 16 Nov 2026, before #1 reached Google; Search Console search appearance + GA4; checks/amp_check/amp_traffic_{pv,dh}_2026-10-20_2026-11-16.json). Clicks Google served as AMP:
| PV | DH | |
|---|---|---|
Web (AMP_TOP_STORIES + AMP_BLUE_LINK) |
240,559 of 3,359,126 (7.2%) | 429,995 of 1,295,268 (33.2%) |
Discover (AMP_TOP_STORIES) |
17,150 of 7,954,006 (0.2%) | 139,534 of 7,485,188 (1.9%) |
| Google News | none | none |
| All three | 257,709 of 11.32M (2.3%) | 569,529 of 9.21M (6.2%) |
GA4 sessions landing on /amp/ |
162,182 of 10.70M (1.5%); 2.7% of Organic Search | 521,258 of 12.42M (4.2%); 6.1% of Organic Search |
- Web share is stable (Sep 2026: PV 7.9%, DH 30.6%;
amp_traffic_{pv,dh}_2026-09-01_2026-09-30.json). AMP is still served for nearly every mobile article click (PV ≈96%; DH's page-level count runs short, so its ratio exceeds 100%: read it as "nearly all"). - Discover AMP is falling on both sites: PV 1.8% (Sep) → 0.2%, DH 5.0% → 1.9% of Discover clicks. Cause not known (no change on our side found); it lowers the traffic at stake if AMP is retired.
- PV's window includes the Deepavali dip from 6 Nov (#42); shares are ratios, so they are less affected than totals.
Decision waits on #1 (confirmed 19 Nov 2026), and why:
- PV: the AMP copy is today the only copy with the full text in server HTML, and Google serves it for ~96% of mobile article web clicks. Retiring AMP before #1 would move that traffic onto pages Google reads without body text (#2: ~76% of PV articles indexed without it).
- Attribution: #1 and an AMP change both shift what Google crawls and serves; doing them in the same window makes neither measurable.
- What #1 changes for this decision: once main pages carry the body, AMP no longer adds content, so the only question left is traffic risk vs crawl cost.
- Timing: DH #1 live 19 Nov 2026 → DH decision from ~17 Dec 2026 (with the #2 re-test), staged by section. PV: ~4 weeks after PV's #1 release.
- Doesn't wait: #40 (no AMP or a paywall for premium stories). The PV print-copy middle ground (option 2 below) also changes AMP, so hold it until PV #1 is live too.
- At decision time re-run:
python -X utf8 checks/amp_check/amp_traffic.py --site <pv|dh> --start <28 days> --end <last GSC day>and compare with the table above (same 28-day length).
What: every PV article exists twice: the main page (indexed) and an AMP copy at /amp/story/<section>%2F<slug>-<id>, announced by <link rel="amphtml"> on the main page. The AMP copy's canonical points to the main page, so it is not indexed as a duplicate, but Googlebot fetches it.
Evidence
- Crawl stats → By response → OK (200), PV, export 12 Nov 2026 (
crawl_stats/2026-11-12/pv-200/): 453 of 993 sample URLs (46%) are/amp/story/…. Sample-based; the true share may differ, but it is large. - PV gets ~360 Googlebot requests per new article vs ~2,700 on DH (#3), so crawl spent on AMP copies competes with main pages and new articles.
- The AMP copy carries the article text in its HTML; the main page doesn't until #1 ships. One article checked live on 12 Nov 2026 (
…-4387960): main page 3 paragraphs / 723 characters in server HTML, AMP copy 6 paragraphs / 958 characters. Google indexes the main page, so the complete AMP text doesn't help search. One article: illustrates, doesn't measure. - Google has given AMP no ranking or Top Stories advantage since 2021; its remaining effects are a second crawled copy of every article and code to maintain (e.g. DH's AMP redirect bug, #14, and DH's AMP 5XX outage 24 Aug – 1 Sep 2026, #3).
Not part of this: PV's amphtml link uses %2F, but on PV that URL returns 200 directly (no redirect, unlike DH before its #14 fix), so changing it to plain slashes saves nothing. ⏸️ Keep it on hold; it becomes moot if AMP is retired.
Plan
- Now (read-only): measure traffic that still arrives through AMP: Search Console search-appearance "AMP" clicks/impressions per site (web, Discover, News), last 3 months; GA4 sessions on
/amp/pages. - Ship #1, wait ~4 weeks, recheck body indexing (#2:
checks/body_index_check/) and crawl (#3: Crawl stats). Don't change AMP in the same window, so effects can be attributed. - Decide: keep AMP, or retire it if AMP traffic is small.
- If retiring: remove the
amphtmllink from article pages; 301 every/amp/story/…URL (encoded and plain forms) to its main page; keep the redirects permanently; check mobile speed of main pages first (#5); watch Discover/News traffic and Crawl stats for 4 weeks. Same for DH if its numbers look similar.
Verify (if retired): amphtml absent from article HTML; AMP URLs 301 to the article in one hop; Crawl stats 200 samples no longer dominated by /amp/story/; crawl of main pages per new article rises.
16 Nov 2026, DH too: in the DH "By file type → HTML" sample (12 Nov 2026, crawl_stats/2026-11-16/dh-html/Table.csv) 491 of 986 HTML fetches (50%) are /amp/story/…; PV 441 of 991 (45%). AMP copies take about half of Googlebot's HTML fetches on both sites.
Why AMP takes ~half of Googlebot's HTML fetches (code + live check, 16 Nov 2026):
- Not a bug or a crawl trap. AMP pages link to the normal article URLs, not to other AMP pages (live: PV AMP article 6 article links / 0 AMP links; DH 2 / 0), and daily sitemaps list no AMP URLs (PV 0 of 590, DH 0 of 129 on 10 Nov 2026). Google finds each AMP copy only through the article's
<link rel="amphtml">and re-fetches it about as often as the main page, which is normal for paired AMP. - Where the link comes from: Quintype
@quintype/seogeneratorStoryAmpTagsin the SEO setup: PVapp/server/app.js:592(generators: [..., StoryAmpTags]) @70d7bebf; DHapp/server/app.js:699@47a7e85e. It addsamphtmlto every story whose template is not inignoreAmpHtmlStoryTemplates(library default: visual story only). PV's link is a relative path (/amp/story/district%2F…); Google resolves it. - AMP page weight: PV ~127 KB, DH ~234 KB per AMP article (live, 16 Nov 2026).
Options (decide after #1 has been live ~4 weeks, as planned):
- Retire AMP: remove
StoryAmpTagsfromgeneratorson both sites; 301/amp/story/…to the article (keep permanently). - Middle ground, PV first step (after PV #1 is live):
ignoreAmpHtmlStoryTemplates: ["visual-story", "syndicated-print"]in the SEO config, so print copies (79% of PV output, little traffic) lose their AMP copy while other templates keep it; 301 their AMP URLs. Roughly removes most PV AMP crawl with one config line.
AMP traffic, measured properly (16 Nov 2026; Search Console search appearance and GA4, 16 Aug – 13 Nov 2026; checks/traffic_funnel/amp_traffic_gsc_2026-11-13.json, amp_traffic_ga4_2026-11-13.json). Clicks Google served as AMP:
| PV | DH | |
|---|---|---|
Web (AMP_TOP_STORIES + AMP_BLUE_LINK) |
946,321 of 12,904,219 (7.3%) | 1,310,275 of 4,495,898 (29%) |
Discover (AMP_TOP_STORIES) |
242,252 of 28,536,333 (0.85%) | 649,435 of 24,550,565 (2.6%) |
| Google News | none | none |
| All three | ≈1.19M of 41.5M (2.9%) | ≈1.97M of 30.3M (6.5%) |
- Page-level
/amp/URL clicks are only 0.20% (PV web) and 0.72% (DH web): Search Console mostly credits AMP clicks to the canonical URL, so the earlier "2.0%" (page-level) under-stated AMP. - GA4, sessions landing on AMP pages: PV 752,063 (1.9% of 39.96M sessions), DH 1,784,868 (4.5% of 39.38M); mainly Google organic (PV 464,795; DH 997,335) and "Direct" (PV 263,270; DH 613,754, likely AMP-cache / Google app visits without a referrer); DH also
googlequicksearchbox100,512 andnews.google.com52,447. - Reading: retiring AMP does not delete these clicks: Google has given AMP no ranking or Top Stories advantage since 2021 and would show the canonical page instead. The risk is the switch-over (canonical pages must be complete and fast). PV: only after #1 (today the AMP copy is the only one with full text). DH: 29% of web clicks are AMP-served, so retire only in stages (one section first) with monitoring. The PV middle ground (no AMP for
syndicated-print) stays the low-risk first step. Severity stays High for the crawl cost (~45–50% of HTML fetches on both sites).
Why DH has more AMP traffic (16 Nov 2026; Search Console web, 16 Aug – 13 Nov 2026; checks/traffic_funnel/amp_share_context_2026-11-13.json): traffic mix, not a technical difference. Only articles have AMP copies, and Google serves AMP to mobile users. PV: 98% of web clicks are mobile but only 10% land on articles (1.53M of page clicks; the rest is home, cartoons, e-paper and listings from branded searches); DH: 85% mobile and 53% on articles (2.28M). Among mobile article clicks, AMP-served clicks are ≈87% on PV (946K of 1.08M) and ≈97% on DH (1.31M of 1.35M) (approximate: search-appearance and page-level clicks are counted slightly differently). So wherever an AMP copy exists, Google serves it for nearly every mobile article click on both sites: retiring AMP moves almost all mobile article search traffic onto the canonical pages, which is why #1 must ship first (PV) and DH must be staged.
AMP health check: no defect stops Google serving AMP (17 Nov 2026; checks/amp_check/amp_check_{pv,dh}_2026-11-17.md). Asked because other vernacular publishers report ~30% of traffic from AMP. Sample: 4 stories per template (top-2 by web clicks + 2 random), DH 62 stories / 17 templates, PV 68 / 17, live as Googlebot plus Search Console URL Inspection.
- Every template except
visual-story(already an AMP Web Story) has<link rel="amphtml">; every AMP URL returns 200 without redirect, canonical = the article, nonoindex, NewsArticle with 1200 px image; official AMP validator: 0 errors on all 122 AMP pages (validation_{pv,dh}_2026-11-17.json). - URL Inspection
ampResult: PV 68 of 68 PASS, DH 59 of 62 PASS (the other 3: 2/dhbrandpr/stories blocked by robots.txt, see #30; 1 with no AMP result yet). Warnings only: "Image size smaller than recommended" on visual stories (both), PVcartoon(3 of 4),nammakshetra-namma-ooru(4 of 4), a few PVtext/syndicated. - Code (PV
70d7bebf, DH22f4c0f0): AMP links only on story pages (@quintype/seoamp-tags.js:5), no template excluded except visual stories, no access/section condition; no AMP→web redirects except the per-storydisable-amp-for-single-storyattribute; no robots/sitemap rule touching/amp. AMP pages carry nomax-image-previewmeta (the library adds it to story pages only,image-tags.js:133); the main page hasmax-image-preview:large. - Reading: the ~30% benchmark is matched by DH web search (29%). PV's lower share (7.3% of web) comes from its traffic mix: only 10% of PV web clicks land on articles (the rest go to home, e-paper, cartoons, listings), and pages other than articles have no AMP. Adding AMP to more pages would not add clicks (no AMP ranking advantage since 2021). Found on the way: premium stories are free on AMP (#40); PV publisher
@idwith no Organization node (#35); DH live blogs have no LiveBlogPosting on AMP either (#34).
Page indexing export (18 Nov 2026; checks/index_coverage/summary_{pv,dh}_2026-11-18.md): "Alternate page with proper canonical tag" is the largest not-indexed reason on both sites and is almost all AMP: PV 1,328,617 pages (95% of a 1,000-URL sample are /amp/story/…), DH 1,255,065 (99%). Expected for AMP (canonical to the article), but it is the size of the duplicate set Google keeps re-crawling. PV's count grew by 114,586 in 85 days (≈ one AMP copy per new story); DH's is flat.