Case study · Helply · SEO + GEO · Aug 2025 – Sep 2026
A new domain. A giant category. 4.9 million impressions in fourteen months.
Helply competes for search terms owned by Zendesk, Intercom, and Freshdesk. In August 2025, helply.com did 226 clicks for the entire month. By July 2026 it was doing 2,484 clicks and 795K impressions a month. Today it holds over 600 top-ten rankings and books demos from organic search every week. All of it while the company rebranded, changed category twice, changed pricing several times, and rebuilt its website.
Full disclosure: our founder is Helply’s Head of Content. This is our own work, shown with real numbers.
15,290
Clicks, Aug 2025 to Sep 2026
4.93M
Search impressions
226 → 2,484
Monthly clicks, start to peak (11×)
27 → 524
Top-10 US keywords, Jan to Sep 2026
Source: Google Search Console, helply.com property, daily data Aug 1 2025 – Sep 19 2026. Top-10 counts from DataForSEO Labs. GA4 corroboration below.
The starting position
Four problems at once. Any one kills most SEO programs.
A brand-new domain
All the equity lived on the old domain: 44.9K clicks and 17.2M impressions across 900+ URLs at an average position of 20.4. helply.com started from zero authority, DA 15 at first audit, with Google refusing to index it. The brief covered three domains: grow Helply and InstantDocs while keeping Groove alive for the migration.A category pivot, twice
Groove sold a simple shared inbox to small businesses. Helply launched as an AI-native support platform for B2B software companies, then in August 2026 widened to any digital business whose support has to scale faster than headcount. Migrating the old content straight across would have imported the wrong buyer either way.A moving product story
Positioning shifted, pricing changed several times, and in September 2026 the whole marketing site was rebuilt on a new stack with new architecture. Content had to rank while the story it told kept changing.A name that wasn’t theirs yet
Searching “Helply” in 2025 surfaced a Webflow template, an unrelated app, and a Tokyo startup. Today the brand query sits at position 1.35 with a 43% click-through rate.
Exhibit A · the strategy memo, dated week one
“Content Strategy For Helply” · August 7, 2025 · author: Bildad
What happened to that ten-keyword target: 524 top-ten keywords fourteen months later. |
The growth curve
100K monthly impressions in 60 days. 795K by July.
The last three full months delivered 6,517 clicks and 2.33 million impressions. The three months before them: 2,620 clicks and 815K. That is +149% clicks and +186% impressions, quarter on quarter.
Monthly impressions · helply.com (Search Console)
View monthly impressions data
Month | Impressions |
|---|---|
Aug 2025 | 3K |
Sep 2025 | 23K |
Oct 2025 | 145K |
Nov 2025 | 307K |
Dec 2025 | 244K |
Jan 2026 | 318K |
Feb 2026 | 313K |
Mar 2026 | 354K |
Apr 2026 | 246K |
May 2026 | 216K |
Jun 2026 | 781K |
Jul 2026 | 795K |
Aug 2026 | 754K |
Sep 2026* | 435K* |
*September is 1–19 only, on pace for roughly 690K. Clicks followed the same curve: 226 (Aug 2025) → 911 (Nov) → 1,308 (Feb) → 1,984 (Jun) → 2,484 (Jul) → 2,049 (Aug). The June–July surge is the migration calendar and comparison cluster landing at once.
Monthly detail · Search Console, helply.com
Month | Clicks | Impressions | Avg position |
|---|---|---|---|
Aug 2025 | 226 | 3,120 | 12.5 |
Nov 2025 | 911 | 307,114 | 8.3 |
Feb 2026 | 1,308 | 312,918 | 9.3 |
May 2026 | 1,106 | 215,614 | 10.5 |
Jun 2026 | 1,984 | 780,698 | 9.7 |
Jul 2026 | 2,484 | 794,950 | 10.8 |
Aug 2026 | 2,049 | 753,605 | 12.7 |
Sep 1–19, 2026 | 1,384 | 434,554 | 10.6 |
The August position dip is explained, not hidden: DataForSEO logged 1,373 newly ranked keywords that month alone, and new queries enter low. Weekly average position has climbed every week since the site rebuild: 11.9 → 10.9 → 10.2.
The build, month by month
From “Google won’t index us” to 524 top-ten keywords.
The weekly reports tell the story better than any summary. The top-ten list grew for five straight months without churn, then kept compounding through two repositionings and a site rebuild. Durable rankings, not lucky spikes.
Top-10 US keyword rankings over time
Date | Top-10 | What was happening |
|---|---|---|
Aug 28, 2025 | 0 | 14 posts live, DA 15, Google refusing to index. Sitemaps resubmitted, Indexing API wired up. |
Oct 24, 2025 | 10 | First #1s: Fin alternatives, Chatbase alternatives. AI-agent tutorials rank within days. |
Nov 13, 2025 | 20 | Cited in Google AI Overviews. Month three. |
Dec 11, 2025 | 26 | December core update dips the whole industry. Diagnosed same day, recovered inside two weeks. |
Jan 22, 2026 | 31 | Eighteen of them ranked #1. |
Mar 5, 2026 | 33 | February clicks up 51% month over month. |
Jun 2026 | 59 | Migration calendar lands. 1,984 clicks and 781K impressions in the month. |
Jul 2026 | 174 | Peak month: 2,484 clicks, 795K impressions. Full technical audit delivered. |
Aug 2026 | 363 | ICP widened beyond B2B. 1,373 new keywords enter the index. |
Sep 2026 | 524 | New marketing site ships. Weekly position recovers to 10.2. Nine demos booked in one week. |
Counts through March are from the weekly Slack reports (US, tracked keyword set). From June they are DataForSEO Labs counts for the whole domain, which is why the scale jumps.
Obstacles cleared
Real programs hit walls. This one hit six.
Indexing refusal
A zero-trust domain Google wouldn’t crawl. Fixed with Indexing API submission and authority-bridging through the old domain’s blog.A negative SEO attack
A spam-backlink campaign suppressed InstantDocs rankings until the toxic links were audited and disavowed. The domain recovered and climbed.Platform bans
Reddit moderation tightened mid-program. The visibility layer pivoted to comment placement, then a vetted agency.A core algorithm update
Google’s December 2025 update dipped every tracked site. Diagnosed against algorithm-tracking data the same day, fully recovered inside two weeks.A rebrand mid-flight
A 900-URL migration executed May–July 2026 without losing the rankings the first nine months had built.A full site rebuild
September 2026: new stack, new architecture, new positioning. Shipped with every one of 307 URLs self-canonical and carrying structured data, a public staging mirror removed, and Core Web Vitals passing on every URL in field data.
The strategy
Seven plays for a domain that couldn’t outrank anyone. Yet.
Every play below traces back to that week-one memo. The work was execution, not improvisation.
Competitor keyword-gap mining
With zero authority to spend, every early target came from a systematic gap analysis of the mid-market support and AI-chatbot ecosystem, modeled on the playbook Chatbase used to add $103K in monthly revenue from a small keyword set. Over 200 scored keyword variations, each tagged by funnel stage before a word got written, anchored on one money keyword: “AI agent for customer support.”
200+ scored targets ★ difficulty mostly under 30 ★ no head-term vanity fights
Highest-intent pages first: alternatives & comparisons
The highest-converting content shipped first: “alternatives” and “X vs Y” posts targeting competitor brand searches, each anchored on Helply’s structural differentiator (no seat fees, AI included) and linked to comparison landing pages and pricing. This cluster is where the #1 rankings piled up.
#1 in the US: Chatbase alternatives ★ Fin AI alternatives ★ LiveAgent alternatives ★ a dozen more in the top five
Riding rising AI queries
Nobody owns a query that’s three weeks old. A deliberate slice of the calendar chased fast-rising AI search demand: agent-building tutorials, model comparisons, training guides. They built crawl frequency and topical authority the commercial pages then borrowed. The top tutorial hit #2 within a week of publishing. A year later the two agent-builder guides are still top-five pages, at positions 7.5 and 7.7.
#2 in week one ★ #1 within days ★ still top-five pages 12 months on
SEO and GEO as one discipline
From week one the program optimized for AI answer engines alongside Google: llms.txt, structured schema, TL;DR blocks, and content shaped around the questions buyers ask LLMs. While the new domain had zero trust, Helply content ran on the old domain’s high-authority blog, so ChatGPT and Perplexity were citing the product months before helply.com could rank on its own. AI Overview citations arrived by November 2025. By September 2026, GA4 reports a distinct AI Assistant traffic channel with its own key events.
llms.txt ★ server-rendered schema ★ authority bridging ★ a 70-post Reddit visibility layer
An editorial system, not ad-hoc drafting
Every piece runs a repeatable pipeline: SERP research, a templated brief, a draft in brand voice, a four-level edit, automated source verification that live-checks every link and statistic, and passes that strip machine-writing tells. That system made 10 posts a week publishable with a tiny team, then scaled to brief and quality-check an external writing team at 30 articles a month.
Research ★ brief ★ draft ★ 4-level edit ★ verify ★ publish
Migration as portfolio management
All 897 legacy URLs with search traffic were audited and bucketed: port same-slug, port renamed, brand redirects, and deliberate retirement of roughly 370 off-topic legacy posts that would have diluted topical focus. The first redirect batch transferred over 1,000 clicks a year of equity before any writing. By late July, 287 legacy URLs were resolving on the new domain.
897 URLs bucketed ★ ~370 deliberate retirements ★ priority calendar by click equity
Measurement wired to revenue
Reporting matured with the program: weekly ranking reports in Slack from October 2025, a full technical audit in July 2026 with every estimate-based finding checked against Search Console, a Looker Studio funnel hub joining GA4 and Search Console in September 2026, and a weekly report that ties search performance to key events and demo bookings by source.
Weekly since Oct 2025 ★ GSC + GA4 + demo attribution ★ live funnel dashboard
Organic is the highest-intent channel.
The traffic converts. Over the trailing twelve months, organic search sessions triggered key events at more than double the rate of direct traffic and engaged for roughly twice as long. In the week of September 7–13, 2026, every demo booked came from Google organic. The following week was the strongest on record: nine demos from four sources.
3.35%
Organic key-event rate (GA4, 12 months) vs 1.59% direct
57.6%
Organic engagement rate, week of Sep 14–20
3 → 9
Weekly demo bookings, Sep 7–13 to Sep 14–20
Third-party corroboration · DataForSEO Labs, Google US
Month (2026) | Keywords | Top-10 | Paid equivalent |
|---|---|---|---|
January | 171 | 27 | $5,131/mo |
April | 179 | 34 | $7,577/mo |
June | 248 | 59 | $12,864/mo |
July | 717 | 174 | $63,856/mo |
August | 1,489 | 363 | $106,177/mo |
September | 2,045 | 524 | $111,447/mo |
Head-term SERPs Helply wasn’t in a year ago · last 28 days, US
Query | Avg position |
|---|---|
best help desk software for small business | 3.4 |
ticketing system for small business | 3.7 |
best ticketing system | 4.2 |
ai agents for small business | 4.6 |
help desk software for small business | 5.1 |
chatgpt agent builder | 5.8 |
helply.com did not enter DataForSEO’s index until early 2026. Nine months later: 2,045 ranked keywords, 524 in the top ten, 15 at #1, and organic visibility DataForSEO prices at over $111K a month in equivalent paid spend, on zero ad budget. The index covers US desktop only, so it understates the global Search Console actuals. Roughly two-thirds of the last 28 days’ 1,882 clicks were non-brand.
Pick fights a new domain can win
Every early target had low difficulty and a beatable SERP. No budget burned attacking head terms like “customer service software.” The rankings stuck, and the head terms opened on their own a year later.Comparisons pay, tutorials build
Alternatives pages converted buyers and stacked #1s. Fast-moving AI tutorials built the authority that lifted them, and are still top-five pages twelve months on.Treat Google and LLMs as one audience
Schema, llms.txt, and brand mentions where LLMs read brought AI Overview citations by month three and a measurable AI-assistant traffic channel by month thirteen.Deciding what NOT to migrate is strategy
Retiring hundreds of off-topic legacy posts protected topical focus. Focus is a ranking asset.One system, any team
The codified pipeline scaled from solo output to an external writing team at 30 articles a month without a quality collapse.Report weekly, against revenue
Weekly reporting against demos booked kept the program funded, caught an algorithm update and a negative SEO attack within days, and now decides which posts get refreshed first.
What’s next
The technical baseline is reset. The remaining debt is editorial.
A 172-post blog refresh runs in four waves, ordered by top-10 keyword count and GA4 key-event rate, to align every post with the widened ICP and current pricing. Title and meta rewrites on the two highest-impression posts (128K impressions between them at 0.1% CTR) are the highest-leverage move on the board. The commercial pages, still under-cited by AI engines relative to the blog, get the next round of structured-data and llms.txt work.
Data sources: Google Search Console (helply.com URL-prefix property, daily export Jul 14 2025 – Sep 19 2026; 28-day page and query export dated Sep 19 2026) · GA4 property helply.com · DataForSEO Labs Historical Rank Overview (Google US, Jan – Sep 2026, pulled Sep 21 2026) · weekly SEO reports in the #seo-geo Slack channel (Aug 2025 – Sep 2026) · weekly performance reports of Aug 31, Sep 14 and Sep 21 2026 · helply.com technical audit (Jul 27 2026) and full-site crawl (Sep 17 2026).
Free · no obligation
Your category has giants too. They write badly.
Send your URL and we’ll find the queries your competitors are fumbling: where you rank, where you could, and what AI answers say about you. In your inbox within a week. Yours to keep, whether or not you hire us.
