How To Do Content Updates On Commercial Pages In 2026 [Backed By 59 SEO Tests]

Content updates on commercial pages are one of the highest-ROI activities in SEO. The below presentation is from my talk at Campixx 2026 in Berlin.

The talk demonstrated exactly how we do content updates at hoppa — and what they actually delivered.

Backed by 59 SEO tests run across 2025 and 2026, this talk walks through the complete 14-step content update process we use for our main moneymaking pages: from GSC diagnostic and KEEP/FIX/REMOVE/ADD tagging, through delta briefs and targeted writing, to SEO preservation and statistical measurement.

The second half covers how we automated the same process using Claude Code — bringing content update time from 8 hours per page down to approximately 30 minutes, without compromising output quality.

The results across 59 tests:

+49% organic click change
+25% aggregate uplift in organic purchases
+20% aggregate uplift in organic revenue

All on commercial pages. All measured with 56-day before/after windows locked in SEOTesting before a single word changed. Revenue and purchase data tracked through GA4 organic attribution.

Transcript

1. C A M P I X X B E R L I N 2 0 2 6 S E S S I O N T A L K How To Do Content Updates On Commercial Pages In 2026 Backed By 59 SEO Tests Alex Galinos Director of Marketing Acquisition · Elife Group (hoppa & Elife Transfer) Elife Group · hoppa.com · elifetransfer.com
2. ME ET T H E S PEA K E R Alex Galinos Director of Marketing Acquisition Elife Group · hoppa & Elife Transfer Ahrefs Blog Semrush Blog Search Engine Land GSA Judge International SEO & Content Marketing Strategist Diverse experience across ecommerce, gambling, finance, health, travel, academic studies and parenting — in 20 markets across the world. Director of Marketing Acquisition, Elife Group Currently managing hoppa & Elife Transfer in the transportation industry. Previously Head of SEO & GEO at the same company. Co-Founder, Skyrocket Growth Premium SEO agency. Considered one of the most influential SEO experts in Greece, Cyprus and the wider region — known for a revenue-focused SEO approach. Co-Founder & Head of Marketing, Parents Hub One of the biggest parenting platforms in Greece. Grew from 0 to 100,000+ monthly users in under 12 months since launch. International Speaker & Industry Contributor Featured in Ahrefs Blog, Semrush Blog, and Marketing Week. Contributor on Search Engine Land. Judge in the Global Search Awards (SEO category).
3. SET TING THE SCENE — WHAT THIS TALK IS ACTUALLY ABOUT hoppa — global ground transportation marketplace. 182+ countries. 2,600+ airports. 9 languages. Thousands of commercial pages. Every single one of them decays over time. For those not from the industry: airport transfer pages are our primary revenue-generating pages — a visitor lands on one, searches, and books. These are the pages that make the business run. This talk is NOT about new pages New pages start at zero — no rankings, no equity, no backlinks You can do whatever you want with a blank slate This is the easy part of SEO to talk about This talk IS about content updates You're working with a live asset that already ranks for something Internal links, schema, historical baseline — all in place The job: refresh what's stale, add what's missing, never break what earns "I've seen plenty of teams rewrite a decaying page from scratch, lose every ranking it had, and call the result a 'refresh.' It isn't a refresh. It's a self-inflicted demotion." Thousands hoppa pages across 9 languages 2 days manual work per page update ~30 min with Claude Code automation
4. THE PROBLEM — CONTENT DECAY IS SILENT, SLOW AND COSTLY Content decay on commercial landing pages costs revenue — quietly and consistently. No error. No penalty. Just gradual relevance erosion. A real hoppa page: Antalya Airport Here's what the 56-day diagnostic showed before any intervention: 148,537 Total Impressions (56 days) 14.89 Avg Position 1.49% CTR 2,215 Total Clicks (56 days) Rankings slip silently A #3 page can drift to #9 — no error, no penalty Competitors refresh · SERPs evolve · you stay still Google notices before your analytics team does Intent mismatch grows New queries emerge that the page doesn't answer AI Overviews and SERP features absorb clicks Same position, falling CTR = losing relevance Revenue bleeds out 148,537 impressions — buried at pos 14.89 Not a visibility problem — a relevance problem Every day unaddressed compounds the loss
5. HOW WE DECIDE WHICH PAGES TO UPDATE Not every page makes the cut. We apply a multi-factor prioritisation framework before a single word is written. 01 High-Priority Business Pages Flagship routes, key markets, homepage-level pages First priority regardless of decay score 02 Pages Showing Content Decay Identified via SEOTesting Content Decay report Click decline, position drift, or CTR erosion not explained by seasonality 03 High Average Booking Value Long-haul routes, premium vehicles, business travel segments Higher ROI per update hour invested 04 Strong Profit Margin Destinations Better supplier margins = traffic gains compound faster SEO effort converts to disproportionate profitability 05 Strategic Expansion Objectives New partnerships, geographies, or competitive gaps Business-identified growth targets get queue priority
6. THE 14-STEP CONTENT UPDATE PROCESS — END TO END Part 1 covers the manual process, walked through a real page. Part 2 shows how the same process runs in Claude Code at batch scale. 01 Read the data 56-day GSC window 02 Tag every section KEEP / FIX / REMOVE / ADD 03 Read competitors SERP audit, new ground 04 Refresh keywords Ahrefs, defended + gaps 05 Rebuild personas GSC clusters + Fan-Out 06 Local knowledge Ops, reviews, current info 07 Decide the angle Thesis for this update 08 Write delta brief Change list, not full brief 09 Write the delta FIX + ADD only 10 Fact-check all New AND kept content 11 Audit images Accuracy, brand, perf spec 12 SEO preservation Preserve unless reason 13 Build UI components Vibe-code the ADD sections 14 Measure the change 56-day baseline test
7. STEPS 1–5 — DIAGNOSE THE PAGE BEFORE YOU TOUCH A WORD Step 1 — Read the 56-day GSC window 56 days — recent enough to reflect current performance without crossing seasonal boundaries. Larger windows mix high and low season data. Clicks dropping — YoY or period-over-period decline on previously strong pages Impressions stagnant — Flat or falling while competitors grow = losing share of voice CTR degradation — Same position, lower CTR — intent drift or SERP features absorbing clicks Position drift — Moving from pos 3–5 to 6–10 — just off the visible fold Query mismatch — New queries appearing but page not ranking = topical gap ⚠ Seasonal caveat: in travel, click drops may reflect off-season patterns. Always cross-check YoY before acting. Step 2 — Tag every section: KEEP / FIX / REMOVE / ADD KEEP Still ranks, still accurate — don't touch it FIX Mostly right but stale — rewrite in place REMOVE Wrong, redundant, or actively hurting ADD Missing — the data says you need it The discipline that separates a surgical update from an accidental rewrite. Steps 3–5 in practice: what the Antalya Airport diagnostic found 20+ striking-distance queries "antalya airport transfer" sitting at pos 7.0 — strong volume, room to move into the top 3. Private transfer gap "private transfer antalya" — 1,091 impressions, 14 clicks, 1.28% CTR. Page had nothing specific for private hire travellers. Comparison gap "best antalya airport transfers" at pos 10.5. No comparison content for users weighing options. 4 distinct personas found Standard shuttle · Private hire · Group traveller · Day-tripper. Built by clustering GSC query data, then imported into SEOTesting.
8. STEPS 6–9 — THE ANGLE, THE BRIEF AND THE WRITING RULES Steps 6–7: Refresh local knowledge, decide the angle What changes in 18–24 months: Terminal assignments Taxi rank locations Scam patterns Tipping conventions Peak congestion Trustpilot themes Antalya's angle: "different travellers need different transfers — surface the right answer based on who you are." Step 8: Write the delta brief — not a brief for the whole page Only what changes gets written. Each section gets a label: KEEP Preserve — defends an existing ranking FIX Rewrite in place — right idea, stale execution REMOVE Cut — wrong, redundant or actively hurting ADD New content — the data says it's missing ~1,500 words per page · reads like an engineering change request, not a writer's brief Step 9: Write the delta — only FIX and ADD sections. The Antalya ADD that mattered: The live "Find Your Perfect Transfer" component — Antalya Airport page Why this works Before: all the information was there — spread across vehicle explainers, group-size guides, FAQs. A visitor had to scan everything to find the bit that applied to them. After: the component is a direct path to information the page already had.
9. STEPS 10–13 — QUALITY GATES, VIBE-CODING AND DEPLOYMENT 10 Fact-check everything — including KEEP sections KEEP sections need verification too — staying doesn't mean still accurate Re-verify: distances, prices, transit times, terminal assignments Most AI-refreshed pages update surface text, skip the underlying facts 11–12 Image audit + SEO preservation Per image: still accurate? on-brand? WebP/AVIF/lazy-loaded? URL slug never changes · meta title left alone if earning CTR Schema extended, never replaced · internal links audited both directions Step 13: Build UI components via vibe-coding Vibe-coding the components Describe visual behaviour in natural language Claude or Gemini generates the React component (tables, carousels, accordions, tabbed components) Iterate live until it works — under an hour for moderate complexity SSR so LLMs can read content without executing JS Old workflow vs now Before: Figma → design review → dev sprint → QA → ship (~2 weeks) After: natural language → working component (~1 hour) Makes per-page components feasible at portfolio scale
10. STEP 14 — MEASURE THE CHANGE Antalya Airport Transfers 56-day test window Metric Control Test Change Clicks / day 39.55 49.00 +23.88% Impressions / day 2,652 2,744 +3.44% Avg position 14.89 10.87 ▲ improved CTR 1.49% 1.79% +0.30pp Queries / day 366 389 +6.28% What drove it Fresh diagnostic · clear angle · delta brief New content written only for the gaps Persona component — direct path to the right info Why this result matters Every metric went the right way. Clicks up · impressions up · CTR up Position improved · queries served up No trade-off — no 'clicks up but impressions dropped' story This is what the test is for. 56-day window locked in SEOTesting before a word changed Baseline separates a measured update from a lucky publish Without it, you can't tell a system from a coincidence
11. WHAT THE DIFFERENCE LOOKS LIKE — TWO REAL HOPPA PAGES hoppa.com/.../austin-bergstrom-international ✗ NOT UPDATED hoppa.com/.../new- york-john-f-kennedy-intl ✓ UPDATED ✗ Generic copy · No local content · No FAQ · No UI components ✓ Rich local copy · FIFA 2026 banner · Trustpilot · Structured sections →
12. PART 2 — HOW THE MANUAL PROCESS BECAME CLAUDE CODE AUTOMATION One manual update = 2 days of focused work per page. hoppa has thousands of pages in 9 languages. The judgment didn't move to the machine — the enforcement of that judgment did. Foundation: we run a dedicated Claude project for hoppa, pre-loaded with our brand book, terms & conditions, pricing policy, and an entire library of brand-specific documents. Every skill that produces content runs inside this project — so brand constraints are always in context, not re-prompted each time. Step 1 hoppa-intelligence 56-day GSC diagnostic + baseline lock 4 metrics · top queries · striking-distance · zero-click Step 2 audit skill KEEP/FIX/REMOVE/ADD tagging runs automatically Classifies each section against query performance Steps 3–4 competitor gap skill Ahrefs: defended queries, gap-close queries, new intents SERP analysis: who's outranking us and what they've gained Steps 5–7 editorial intelligence Persona re-validation · Fan-Out gap detection Hard gates: update cannot proceed without angle + local knowledge Step 8 delta brief skill Generates delta-shaped briefs — KEEP/FIX/REMOVE/ADD explicit Will not produce a brief without a defined angle Step 9 hoppa-editorial Writes FIX and ADD sections only, inside the hoppa Claude project Re-fires if tone metrics diverge from gold-set benchmarks Step 10 hoppa-scientific-refiner Fact-checks new AND kept content Failed checks move KEEP to FIX automatically Step 11 image-auditor Accurate? On-brand? WebP/AVIF/lazy-loaded? Flags replacements · generates missing alt-text Step 12 seo-preservation URL slug locked · meta title preserved if earning CTR Schema extended, never replaced · internal links audited Step 13 component generator Generates React components from brief specs Describe behaviour in natural language · iterate live
13. WHAT RUNNING THIS LOOKS LIKE — AND WHERE IT'S HEADING In practice: a batch of up to 30 pages 1 We set the list Revenue weight + strategic picks. The list is ours — the system runs on the pages we bring. 2 Batch runs unattended Diagnostic → inventory → gap analysis → keyword refresh → persona re-validation. All unattended. 3 Delta briefs generated Writing, fact-checking, image audit, component generation, SEO preservation, staging deployment. All automatic. 4 We review and approve Up to 30 updated pages. ~30 min of review per page. Then publish. Where this is heading Today: revenue picks first We defend the pages revenue says matter. Correct, but reactive — decay has already cost us something by the time it surfaces. Tomorrow: the system picks first CTR softening, position drift, query-coverage gaps — the signals exist before revenue moves. The roadmap: a system that watches the full portfolio and flags pages before they bleed. What we're deliberately not sharing The prompts, gate criteria, persona schema, voice benchmarks, tagging rules. Rebuild all of it for your domain. The architecture transfers. The specifics don't.
14. WHAT TRANSFERS TO YOUR TEAM — THE HALF THAT MAT TERS Throw out everything specific to hoppa. The vertical, the personas, the voice, the preservation rules — rebuild all of it for your domain. What transfers is the architecture. 01 Human direction at the front You set the list, the priorities, and approve the output Automation enforces judgment — it doesn't replace it 02 Hard gates that enforce judgment in the middle Skip persona validation, local knowledge, angle decision → Result: generic AI — it's why every pipeline sounds the same 03 Preservation discipline protects what's earning Breaking rankings isn't an update — it's a demotion KEEP sections are untouchable · work with the asset 04 Measure against the page's own baseline 56-day windows locked before a word changes 'Did this change move the baseline?' not 'how is it doing?' 05 A loop that feeds learning back into the next batch Antalya +24% with every metric right Each result becomes a prior for the next decision
15. 59 SEO TESTS — THE AGGREGATE RESULTS Every test has a 56-day before/after window locked in SEOTesting before a single change ships. These are the real numbers. 59 SEO Tests run all completed 42 Organic clicks improved 71% of pages 15 Organic clicks declined 25% of pages 2 Flat (no change) organic clicks +49% Median organic click change across 59 tests +129% Mean organic click change skewed by top outliers Clicks Change — Distribution Best result single page +1,167% Strong wins (>100%) 25% of all tests 15 tests Wins (50–100%) 22% of all tests 13 tests Wins (<50%) 24% of all tests 14 tests Losses worst: −45% 15 tests Absolute Click Totals — All 59 Tests Normalised to 56-day windows for comparability 10,195 Clicks before control period total 12,479 Clicks after test period total +2,284 Net uplift additional clicks Every one of these clicks landed on a transactional transfer page — not a blog, not an article. Each additional click is a potential booking. Absolute Organic Click Totals — All 59 Tests Normalised to 56-day windows for comparability 10,195 Clicks before control period 12,479 Clicks after test period +2,284 Net uplift additional clicks Every click landed on a commercial page, not a blog. Each additional click is a potential booking. Organic Sales aggregate uplift +24.9% organic purchases content updates vs baseline Organic Revenue aggregate uplift +20.1% organic revenue (£) content updates vs baseline
16. Q&A Thank you for listening. Alex Galinos Director of Marketing Acquisition — Elife Group One takeaway from today: Content updates on commercial pages are not maintenance — they are one of the highest-ROI SEO activities. Preserve what earns. Fix what's stale. Add what's missing. Measure everything. The system only works because it has enough context and every test gets a clean baseline. That's how you tell a system apart from a lucky strike. Elife Group · hoppa.com · Campixx Berlin 2026 Special thanks Yvette Ramirez — Content Specialist, Elife Group Responsible for building the Claude Code implementation and skill files that turned this manual process into an automated system.