TheProjectSEO · Executive Report

Helpling Singapore — SEO Recovery: What's Working, What's Broken, What's Next

A single, real-data view of the March 2026 organic decline, the root causes found so far, what's already fixed, and what's still open. Skim the summary below — every section can be expanded for the full evidence.

Reporting window: Jan – Jun 2026, with 16-month trend context
Prepared: July 2026
Sources: Live GSC + GA4 APIs, Ahrefs, BigQuery lake — no modeled/estimated figures
-31.2%
GA4 organic sessions, Jan peak → May low (corrected for a data-integrity issue found this cycle)
Confirmed decline
1,152
Toxic/suspect backlink sources found, across 5 structurally distinct clusters
Active campaign
~5.9x
Growth in AI-platform (ChatGPT/Perplexity/Gemini) referral traffic, Mar 2025 → Jun 2026
Bright spot
0
Of 5 core keywords show a demand drop timed to the decline — this is a visibility problem, not a demand problem
Root-caused

The big picture, in one paragraph

Organic clicks fell hard starting April 2026 (17,899 → 13,181, -26%) and haven't recovered — but demand didn't drop, reviews stayed excellent, and AI-platform referral traffic kept growing right through the decline. That combination points squarely at trust and technical signals, not the business or the market. Three concrete, independently-confirmed causes were found this cycle: a live negative-SEO backlink campaign that's still adding domains today, a physically-removed Product schema on one of the site's two highest-traffic pages, and a robots.txt error that's been silently blocking 93% of Google's crawl access for twelve weeks. Meanwhile, competitor Urban Company gained sharply on every core keyword checked in the exact same window — this isn't a shared industry dip, it's active ground being lost. The fixes are largely built and ready; what's outstanding is mostly sign-off and deployment, not further diagnosis.

Mar 2025
Baseline period — healthy rankings, steady traffic
Mar 2026
Google Spam/HCU update rolls out — site-wide query decline begins
Apr 2026
robots.txt breaks (still live today) + Cluster 1 backlink campaign begins
Feb–May 2026
Product schema quietly removed from /cleaning-services
Jun 2026
Clicks and position start stabilizing; GA4 property found to be globally-contaminated, now corrected
Jul 2026
Full 3-phase diagnosis complete — fixes built, most pending sign-off/deploy

What's working — protect and build on these

Three real, verified bright spots found this cycle. None of these are hypothetical — all are backed by live data pulls.

Growing

AI-platform referral traffic, up ~5.9x

Real human sessions from ChatGPT, Perplexity, and Gemini grew from 93/month to 549/month over 16 months — and kept growing right through the decline. Landing mostly on the homepage, /maids, /elderly-care, and /aircon-services.

Confirmed

Demand and brand search are both healthy

Search volume for every core category term held flat-to-stable, and both brand terms ("helpling", "helpling singapore") grew ~30% over the window. Reviews stayed at 4.6–4.93 stars throughout.

New content working

Aircon informational content is a real gainer

New content like the aircon dry-mode guide and Mitsubishi warranty pages are among the biggest query gainers on the whole site — a template worth repeating for other categories.

What needs fixing — in priority order

Five items, each independently confirmed with real data. Two are already fully built and just need sign-off.

Priority 0

robots.txt returning a server error, 12 weeks running

app.helpling.com.sg — the subdomain carrying 93%+ of crawl volume — has been silently failing since mid-April. Fastest, cheapest fix on this whole list.

Priority 0

Product schema physically removed

/cleaning-services lost its Product schema between Feb–May 2026, confirmed two independent ways. Plausibly explains a meaningful share of a ~15%-of-clicks rich-result collapse.

Ready to deploy

Toxic backlink campaign, still active today

1,152 toxic/suspect sources across 5 clusters — two of them still adding new domains right now. Disavow file is built and reviewed, needs sign-off to submit.

Ready to deploy

Redirect map + PQI cleanup

24+ page consolidation, 12 confirmed 410s, and a technical pagination bug are all built and QA'd. One content sign-off (Singlife/Supercare pages) is the only blocker.

Competitive alert

Urban Company gaining ground fast

Climbed from position 14→2, 7→2, 20→5, and 3→2 on core keywords — in the exact window Helpling fell. Not a shared industry dip; direct displacement.

Traffic — the real, corrected picture

GSC clicks are ground truth throughout. GA4 sessions required a correction this cycle — the property blends in Germany/France/Switzerland/Australia traffic, which reversed the apparent trend direction until filtered to true Singapore hostnames.

Monthly GSC Clicks & Impressions

Live Search Console API, calendar-month aligned

ClicksImpressions (scaled)

GA4 Organic Sessions: Before vs. After Correction

The uncorrected line blends in 4 other countries' traffic

Uncorrected (global blend)Corrected (true SG)

Revenue & the AI-referral bright spot

Average Daily GMV, EUR

Helpling Customer Insights — SEO channel vs. all-channel total

SEO channelAll channels (÷10 scale)

AI-Platform Referral Sessions / Month

ChatGPT + Perplexity + Gemini, true-SG hostnames, real human traffic only

Core keyword performance

Live GSC data, June 2026. The aircon row is the one clear, fixable problem: two of Helpling's own pages compete for the same query.

KeywordRanking pagePositionClicksImpressions
cleaning services singapore/cleaning-services3.61013,202
home cleaning services singapore/cleaning-services4.710903
post renovation cleaning/post-renovation-cleaning4.5341,209
aircon servicing/aircon-servicing-guide (blog — shouldn't be winning)8.9252,902
aircon servicing/aircon-services (commercial page)18.872,741
sofa cleaning singapore/sofa-cleaning4.3111,187

Helpling vs. competitors — position change, Feb → Jul 2026

Ahrefs organic-keyword tracking. Urban Company is the only competitor gaining ground; Sureclean and Cleanlab stayed flat or weak.

KeywordHelplingUrban CompanySurecleanCleanlab
aircon servicing3 → 127 → 2——
home cleaning services sg2 → 53 → 27 → 1927 → 22
cleaning services sgfell out of range14 → 2new: 28new: 21
post renovation cleaningfell out of range20 → 533 → 3534 → 38

Deep dive — full evidence by phase

Everything above is the summary. Expand any section below for the full methodology, data, and reasoning behind each finding.

Phase 1 — Toxic Backlink Audit (1,152 sources, active campaign)

Backlink cluster sizes

What was found

  • Cluster 1 — "SEOExpress" fake-testimonial network: 162 domains, active and adding new ones daily since April 20, 2026. Disposable TLDs, fake first-person case studies naming a business the founder has never heard of.
  • Cluster 2 — "Power up helpling" network: 39 domains, active June–July 2026, uniform DR ~2, all flagged spam by Ahrefs.
  • Cluster 3 — Chinese-anchor network: 84 domains, a single historical burst in 2023, dormant since.
  • Cluster 4 — Raw IP bot/scraper network: 818 IP addresses found only in Google's own Links report, zero overlap with Ahrefs' data. Needs URL-level (not domain-level) disavow entries — a separate decision on whether it's worth the effort given likely transience.
  • Cluster 5 — Gibberish/DGA domains: 49 disposable domains (.forum TLD cluster + word-salad .com names).

Status

  • Manual Actions and Security Issues in GSC both show "No issues detected" — no formal Google penalty yet, but that doesn't rule out algorithmic impact.
  • A submittable, Google-format disavow file (334 domains, Clusters 1+2+3+5) is built and reviewed. Not yet submitted — awaiting sign-off.
  • Broken backlinks pointing at a dead blog subdomain (real DR 38-79 link equity) have been merged into the redirect map already.
Phase 2 — Site-Wide Technical Recovery (schema, crawl, redirects, PQI)

The schema regression

  • /cleaning-services — one of the site's two highest-traffic pages — had its Product schema physically removed between Feb 7 and May 11, 2026, confirmed via a Wayback Machine diff of the page's actual structured data.
  • Independently re-confirmed by Ahrefs' own Site Audit crawl, which also flagged a real Google rich-results validation error on /spring-cleaning and /raya-spring-cleaning-bundle's Review schema.
  • Product Snippets + Review Snippets together were worth ~15% of baseline organic clicks and have collapsed to near-zero.

Technical crawl issues

  • app.helpling.com.sg's robots.txt returns a live HTTP 500 error, confirmed by direct request — broken since ~mid-April 2026, ~12 weeks. This subdomain carries 93%+ of total crawl volume.
  • blog.helpling.com.sg is functionally dead — 83% of Google's crawl requests to it return 404.
  • A blog pagination bug (bare-number URLs duplicating the real paginated archive) plus 3 mistitled category slugs were fully diagnosed and added to the redirect map.

PQI cleanup (page-quality classification against real 16-month GSC data)

Page inventory breakdown

  • Of 22 real "Dead" page candidates checked individually: 7 needed zero action (already resolved or misclassified), 12 are clean, confirmed 410 candidates, 1 needs a content decision.
  • 96 "Weak" pages ranked by real search demand for renovation priority — top candidates already have incoming backlink equity queued from the Phase 1 redirect merge.

Redirect map status

  • 24+ row consolidation map fully built and QA'd. All three original pre-production blockers are resolved or narrowed: noindex re-scrape done, handyman/upholstery decided as content hubs (not deletions, based on real traffic data), only the Singlife/Supercare/Reminisce content sign-off remains open.
Phase 3 — Revenue Growth Layer (demand, reviews, AI-search)

Search demand: 2026-to-date average vs. 2025 average — 5 keywords + 2 brand terms

100% = no change. Category terms use Mar 2025 vs. Jul 2026 endpoints (their trend is directional and stable); *brand terms use full-period averages since single-month comparisons are too volatile to trust on their own.

Demand vs. visibility

  • None of the 5 core category keywords show a demand drop timed to March 2026 — two never declined at all, the two that did had already bottomed out by late 2025.
  • Both brand terms ("helpling", "helpling singapore") stayed within a noisy but stable range — no sustained decline in either. (Single-month comparisons can show +30% or -11% depending on exactly which months you pick; a fairer full-period-average comparison puts both within ±8% of flat. The honest read: brand search held up, this isn't a growth story either way.)

Reviews & reputation

Google Business Profile — monthly reviews & rating

  • Rating held 4.6–4.93 throughout the entire decline window; review volume was elevated, not falling. Reputation is ruled out as a contributing cause.
  • One unexplained one-month spike to 899 reviews in February 2026 — ratings stayed high through it, worth a quick data-sanity check but not concerning.

AI-search visibility

  • Real citation baseline confirmed: Google AI Overviews (82 citations/39 pages), Google AI Mode (65/38), Grok (44/24), ChatGPT (38/28), Gemini (17/13), Perplexity (17/9), Copilot (16/10).
  • GA4 referral data confirms real human traffic behind those citations, growing ~5.9x over 16 months, landing mostly on the homepage, /maids, /elderly-care, /aircon-services, /pricing, /cleaning-services.
  • Gap found: the Brand Radar competitive tracking report (vs. Urban Company) was configured in February 2026 but every data source was left switched off — five months of an inactive tracker.

Category revenue reality check

  • Air Conditioning: fastest-growing category YoY (+49% to +104% every month since February), but organically almost invisible (~€42/day) — the clearest, most quantifiable demand-vs-capture gap on the whole account.
  • Elderly Care: down 33-51% YoY every month, with zero organic GMV attribution in the entire dataset — a category-health question for Helpling's team, not something SEO alone can move.
  • Move out Cleaning: the clear organic standout, #1 or #2 organic category every month, recovering (not yet fully back to 2025 levels).
Data-quality findings from this cycle (11 issues surfaced, each tagged)
  • GA4 property is a global rollup, not SG-scoped — Germany's tracking was merged in around Jan 2026, reversing the apparent session trend until corrected. Fixed in this report; standing filter rule now in place for future pulls.
  • BigQuery ga4_sessions table disagrees with the live GA4 API by ~2-3x — traced to paid-campaign landing pages contaminating the ingestion pipeline. Live API used instead throughout.
  • GA4 revenue tracks only a fraction of real bookings — most orders complete via the app, not the web checkout GA4 sees. Customer Insights used as the source of truth for GMV instead.
  • Customer Insights GMV column stopped populating after June 5-7 — orders kept flowing normally; this needs a fix from whoever maintains that sheet.
  • Google Ads data isn't in the lake at all for this client — access blocker, not a data problem.
  • Ahrefs' crawl-based traffic/position estimates diverged meaningfully from GSC ground truth this cycle — live GSC/GA4 pulls are now the default for this client.