Prepared 21 September 2026 by Vinson Leow, updated the same evening with Search Console and GA4 connected. Based on Search Console (domain property, data from 27 Aug), GA4 (property 551826398, data from 1 Sep), a live crawl of every page in the sitemap, the Hypertune analytics platform (first-party tracker, 28 days to 21 Sep), a link check of the 108 September press pieces in the agency's tracker, the live schema and robots files, and the Fluenta AI-visibility probes (15 Sep). Not yet available: a backlink index (Ahrefs Webmaster Tools, free, is the next step).
Search is already Hypertune's largest converting channel and nobody is managing it. In the last 28 days 14,357 people arrived from a search engine as their first visit and 531 of them became paying subscribers (3.7%), more than YouTube on the same first-touch basis (8,621 visitors, 209 subscribers, 2.4%). Almost all of that is people typing "hypertune" after seeing a creator video. Search Console for 22 Aug to 19 Sep: 8,213 clicks on 107 branded queries, 254 clicks on 1,299 non-branded queries. GA4 agrees from the other side: since the purchase event started firing on 12 Sep, Organic Search accounts for 156 of 312 recorded purchases, half of everything GA4 can see.
The site has 13 indexable pages. Hone, the main competitor, has 1,704: 42 game pages, 292 system-requirements pages, four comparison pages, two research pages, translated into 25 languages. That gap, not technical health, is why AI assistants name Razer Cortex, Afterburner and GeForce Experience for "best game booster" and never Hypertune (0 of 5 engines, Fluenta probe).
The cheapest authority win is sitting in September's press run. Of 108 pieces checked, 12 link to hypertune.gg with a normal link, 5 with nofollow, and 74 name Hypertune without linking at all, including VideoCardz, TechPowerUp, TweakTown, KitGuru, GamesBeat, MCV and Sportskeeda. Asking those outlets for a link is a week of emails, not a month of outreach.
Technical SEO is fine. Fluenta's 78 of 100 readiness score is fair. There are a dozen on-page defects (H1 structure, thin guide titles, a game page returning a server error) that take Kevin and Kamil about a day together.
Neither report on the table is worth paying for as written. Bobby's plan proposes building schema, FAQ answers and a guides section that are already live, has no targets, and asks for $3,000 a month. Fluenta's is an automated lead magnet: its backlink data is from 14 July (before the press run), it calls the Romanian press pickup "spam", and its suggested copy is template output. Both contain a few true things, which are folded in below.
Plan: 90 days, most of it done in house. Outsourced: link outreach at $500 to $1,000 a month from week 4, optionally a hardware-literate copy editor. Details in section 9.
| First visit came from | Visitors | New paying subs | Conversion |
|---|---|---|---|
| Google search | 10,884 | 379 | 3.5% |
| Direct / untagged | 8,670 | 326 | 3.8% |
| YouTube | 8,621 | 209 | 2.4% |
| Other search (Bing, Brave, DuckDuckGo, Yahoo, Yandex) | 3,473 | 152 | 4.4% |
| Other referrers | 3,399 | 44 | 1.3% |
| AI assistants (ChatGPT, Gemini, Perplexity) | 34 | 3 | 8.8% |
Method: first entry page per visitor in the window, matched to accounts that signed in on the same device, then to a subscription created in the window. Visitors who never sign in on that device are not matched, so the absolute counts are floors; the ranking between channels holds. The dashboard's channel table uses a different method (tagged link clicks) and shows YouTube at 395; the two are not comparable and both are right for what they measure.
Where search visitors land: 16,951 of about 18,500 search entries land on the homepage. The guides receive effectively no search traffic (/guides/xtu-bios 392 sessions in 28 days, but from the app and support links, not search, and 91% bounce). /game/valorant had 5 sessions. Nothing but the homepage ranks.
The property was verified on 27 Aug, so there is no earlier history and no month-on-month comparison yet; the 16-month back-fill Bobby's plan mentions does not exist for this property.
| Queries | Clicks | Impressions | |
|---|---|---|---|
| Branded (hypertune and misspellings) | 107 | 8,213 | 17,899 |
| Non-branded | 1,299 | 254 | 7,548 |
| Total (page-level, includes anonymised queries) | 9,913 | 39,000 |
Top non-branded queries by impressions: "pc optimization" 431 impressions, 12 clicks, position 8.2 · "pc optimization for gaming" 369, 30 clicks, position 6.8 · "optimize pc" 247, 2 clicks · "best pc optimizer for gaming" 177, 5 clicks, position 8.4 · "pc optimizer" 170, position 26 · "pc optimizer for gaming" 161 · "computer optimization" 142, no clicks · "cpu game optimizer" 111 · "pc tuning" 107, position 23 · "pc game optimizer" 70, position 19. The homepage is already on page one for the "pc optimization for gaming" cluster with nothing targeting it; a title and opening paragraph aimed at that phrase is the fastest non-branded gain available.
Daily branded clicks ran about 300 from 27 Aug to 8 Sep, then 724, 1,212 and 762 on 9, 10 and 11 Sep, back to about 380 by 15 Sep. That three-day spike coincides with the press embargo lifting on 9 Sep; creator videos in the same week make the split imprecise, but the shape is a news spike, not a creator curve.
Pages: the homepage takes 9,690 of 9,913 clicks. /onboarding?trial received 7,280 impressions and 70 clicks at position 1.5, meaning Google is showing the trial checkout as a sitelink for brand searches even though the page is now noindex; /expert-tuning 13,343 impressions but 46 clicks (0.3%); /game/valorant 425 impressions, 1 click; the five guides together under 100 impressions. Countries: USA 23% of clicks, then UK, Germany, Spain, Australia, Canada, Brazil, Russia, Turkey, India. Devices: desktop 76%, mobile 24% of search clicks (search skews more mobile than the site's 9% overall).
No sitemap is submitted in Search Console (the sitemaps list is empty). Indexing per Bobby's earlier pull: 16 indexed, 370 not, mostly framework chunks and portal routes.
| Channel | Sessions | Sign-ups | Checkouts started | Purchases (from 12 Sep) |
|---|---|---|---|---|
| Organic Search | 9,894 | 1,276 | 1,357 | 156 |
| Direct | 6,410 | 236 | 262 | 31 |
| Organic Video (YouTube) | 5,444 | 607 | 642 | 82 |
| Referral (press and others) | 2,458 | 35 | 99 | 16 |
| Organic Social | 1,434 | 26 | 27 | 2 |
| Unassigned | 772 | 153 | 174 | 20 |
| AI Assistant | 49 | 4 | 5 | 1 |
Two measurement defects, both for Kevin: the purchase key event recorded nothing until 12 Sep and since then logs about 35 a day against roughly 56 a day in Stripe, so GA4 sees about 60% of purchases (probably the pixel-loading change, or a checkout path that never fires it); and 772 sessions are Unassigned with a 5% engagement rate, which is usually a tag firing on a redirect page. Neither changes the channel ranking; both need fixing before Google Ads relies on this property for conversions.
"game booster" 135,000 a month, "fps booster" 33,100, "input lag" 8,100, "pc optimization" 5,400, "gpu overclocking" 5,400, "pc optimization software" 2,400, "msi afterburner alternative" 880, "razer cortex reviews" 140. Difficulty is low on most of these (KD 0 to 31). Volumes need confirming once we have our own keyword tool, but the shape is clear: the category is large, the brand is invisible in it.
Working correctly: robots.txt allows everything and points to the sitemap; sitemap lists 13 URLs and all return 200; app.hypertune.gg and www both 301 to the bare domain; unknown URLs return a real 404; /onboarding, /sign-up, /download and the portal are noindex, as they should be; Open Graph tags on every page; existing JSON-LD (SoftwareApplication with publisher Organization, and a 7-question FAQPage) parses cleanly. Speed work already on the test site takes blocking time from about 1 s to 101 ms.
Defects:
| # | Finding | Why it matters | Owner, effort |
|---|---|---|---|
| T1 | /game/cs2 and /game/fortnite return HTTP 500, not 404 | Server errors on crawlable paths waste crawl budget and look like a broken site to Google; if a game-page template exists, only Valorant is wired up | Kevin, 1 h |
| T2 | Homepage has no canonical tag (every other page does) | With www and ?ref= variants in the wild, Google picks the canonical itself | Kevin, 15 min |
| T3 | Sitemap has no lastmod dates | Google uses lastmod to decide what to recrawl; without it, changes to the homepage wait for the normal cycle | Kevin, 30 min |
| T4 | /technology has 6 H1s, /expert-tuning 8 (three of them are "0+", "1ms", "673 FPS" counters), /game/valorant 5 | H1 is the page's topic statement; six of them is none | Kamil, 1 h |
| T5 | All five guides and /guides have no H1 at all | Same problem the other way | Kamil, 30 min |
| T6 | Guide titles are 9 to 24 characters ("XMP Guide", "AMD BIOS Guide") with no brand and no target phrase | The title is the search result headline; these cannot rank for anything | Vinson writes, Kamil pastes, 1 h |
| T7 | Guides are 126 to 429 words each; /guides index is 77 words | Thin pages do not rank and are not cited | Content plan, section 5 |
| T8 | /game/valorant is in the sitemap but not linked from anywhere on the site | Orphan pages rank poorly; it also duplicates the homepage structure (two "Introducing Hypertune" H1s) | Kamil, 30 min |
| T9 | Meta description on /policy is 350 characters; /guides 69; /terms 93 | Truncated or empty snippets; minor | Vinson, 20 min |
| T10 | 5 to 18 images per page with no alt text (144 images on the Valorant page, 18 without) | Accessibility and image search; minor | Kamil, 1 h |
| T11 | Member guides are duplicated at /client-portal/member-guides/* | Portal is noindex so no penalty today; one source of truth is still cleaner | None now |
| T12 | No llms.txt | Fluenta flags this; it is a convention some AI crawlers read and Google ignores; 20 minutes, low value, harmless | Kevin, 20 min |
| T13 | Google's index still holds old homepage copy promising a "full refund, no questions asked" (Bobby's finding) | Request re-indexing after the on-page changes ship | Vinson, 10 min in GSC |
Fluenta's crawler probe was blocked by the edge (IP and TLS filtering), so it could not confirm that Googlebot and GPTBot get the same page a browser does. Worth one check in GSC's URL Inspection (live test) and one fetch with the GPTBot user agent once access is in place.
| Page | Title today | Words | Verdict |
|---|---|---|---|
/ | Hypertune, Ultimate PC Optimization for Gaming & Esports | 1,032 | Targets the brand only; carries the whole business; rewrite title and description around one non-branded phrase plus brand |
/technology | Five features behind one click | 681 | Title says nothing a searcher types; this page should own "how PC optimization software works" or similar |
/expert-tuning | Handtuned by Experts | 923 | Only page with a clear service; title should say what it is (remote PC tuning service) |
/tutorial | Two minutes away from more FPS | 309 | Thin; fold into onboarding or expand |
/guides and 5 guides | XMP Guide, AMD BIOS Guide, Intel XTU Guide, BIOS Update Guide | 77 to 429 | Real search demand exists for every one of these ("how to enable XMP", "intel xtu guide"); none rank because they are thin and untitled |
/game/valorant | Tune your PC for Valorant in one click. | 1,764 | The best page on the site by structure and the model for the game series; orphaned and duplicated headings |
/policy, /terms-of-service | fine | Nothing to do |
Bobby's on-page recommendations (one H1, unique descriptions, keyword per page, alt text) are correct and standard. Two corrections to his document: SoftwareApplication and FAQPage schema are already live, and both of his "first-pass FAQ examples" are already in the live FAQ. The schema work that remains is additive: offers (price), aggregateRating once reviews exist on-site, sameAs to the social profiles, an explicit top-level Organization, and real titles on the VideoObject entries.
FAQ rich results: Google withdrew FAQ dropdowns from all but government and health sites in August 2023. FAQ schema still helps machines read the page; it will not produce expandable results for Hypertune. Bobby's document says otherwise.
Hypertune has 13 pages. Hone has 42 game pages and 292 system-requirements pages in English alone, plus comparison pages ("Hone vs X") and research pages, in 25 languages. AI assistants and Google both answer category questions from sites that cover the category. Hypertune covers itself.
What to build, in priority order, using data we already own:
/game/valorant exists as the template. Next titles come from two sources we have and Hone does not: game mentions in Discord and support tickets (filtered to real games), and the creator videos that already convert. Likely list: CS2, Fortnite, Apex Legends, Warzone, Marvel Rivals, Rust, Overwatch 2, League of Legends, Rainbow Six Siege, plus Valorant fixed. Each page: what the game is bound by (CPU, GPU, network), the settings that matter, what Hypertune changes, a measured before-and-after from our own benchmark data for that game if available, an FAQ. 1,200 to 1,800 words each. This is also what the paid creator campaigns should link to instead of the homepage.Not to build: the adjacent-interest content in Bobby's plan (anime, music, fashion, fitness). It dilutes topical authority and there is no traffic case for it.
| Status | Count | Examples |
|---|---|---|
| Links to hypertune.gg, normal link | 12 | Tom's Hardware, PurePC, Tarreo, Basic-Tutorials (EN and DE), SoftZone, HardwareCooking, Connect.ro, Știrile pe scurt |
| Links with nofollow | 5 | HWBusters, Club386, Overclocking.com, Tom's HW Italy, Yahoo ES |
| Names Hypertune, no link | 74 | VideoCardz, TechPowerUp, TweakTown, KitGuru, GamesBeat, MCV, Sportskeeda, 3DCenter, Blue's News, HardwareBrief, ITHardware.pl, Profesional Review, and 60 more |
| Blocked our check or not about Hypertune | 17 | Igor's Lab, HWUpgrade, Adrenaline, Cowcotland (bot-blocked; check by hand) |
The full list with an action per row is in Press coverage backlink check 2026-09-21.csv. Seventy-four outlets already wrote about the product; asking each for a link to the page they are describing is normal practice and converts at perhaps 20% to 30%. That alone would triple the referring-domain count Fluenta measured (8 on 14 July).
Fluenta's "your backlink profile is a liability" and disavow file: wrong. The one domain it flags, stirilepescurt.ro, is a Romanian tech outlet that covered the launch. Do not upload the disavow file.
Not recommended: HARO-style quote farming, Medium listicles, Slack link-building groups, buying links of any kind. The first two produce little; the last risks the domain.
Where it stands: ChatGPT, Claude and Grok do not name Hypertune for brand or category questions; Perplexity and Google AI Mode cite hypertune.gg for brand questions only (Fluenta, 15 Sep). Category recall 0 of 5. AI referrals in our data: 82 sessions from ChatGPT, 6 from Gemini, 3 from Perplexity in 28 days, converting at a high rate on a tiny sample.
Why: AI assistants draw on the same signals as search, plus a bias toward pages that state facts plainly and toward brands that appear in third-party comparisons. Hypertune has no comparison presence, no category content, and a thin link profile. The technical layer is already adequate. Fluenta's own summary is right on this point: the fix is not the website.
What moves it, all already in the plan above: the comparison pages (section 5, item 4), the methodology page (item 3), the press links (section 6), directory listings with consistent facts, and Bing Webmaster Tools, because ChatGPT search reads Bing's index and Bing is not yet verified. Additional, small: an llms.txt, and Organization schema with sameAs.
Measurement, free: a fixed set of ten prompts ("best PC optimizer for gaming", "is Hypertune safe with Vanguard", "Hypertune vs Hone", and so on) asked of ChatGPT, Claude, Gemini, Perplexity and Grok on the first of each month, answers recorded in a sheet. Surfer's AI Tracker at about $95 a month is not needed until the count is above zero.
Connected 21 Sep: Search Console (domain property, full user) and GA4 (Analyst) through a Google Cloud project hypertune-seo-data owned by vinson@cosmicwire.com, plus the analytics platform, Klaviyo and the affiliate database. Both Google sources are queryable from this workspace (scripts/gsc.sh for Search Console; the Google Analytics connector in the project config).
Still to do, all small:
Monthly report, one page: branded vs non-branded clicks and impressions, indexed pages, referring domains, search-first-touch visitors and subscribers from the platform, the ten AI prompts, top guide and game pages by entries.
| Phase | When | Work | In house | Outsource | Hours |
|---|---|---|---|---|---|
| 0. Access and baseline | This week (GSC and GA4 done 21 Sep) | Submit sitemap; Bing WMT; Ahrefs WMT; ten-prompt AI baseline; non-indexed URL list from GSC | Vinson 3 h | None | 3 |
| 1. Fix and ask | Weeks 1 to 3 | T1 to T13; titles and descriptions for all 13 pages; schema additions; VideoObject titles; press link requests to 74 outlets; directory listings; request re-indexing | Vinson 16 h writing and emails; Kevin 1 day; Kamil 1 day; Austin 1 h sign-off | None | ~40 |
| 2. Build | Weeks 2 to 8 | 10 game pages; safety page (with sign-off); methodology page; 4 comparison pages; 5 guides rewritten; review flow in Klaviyo | Vinson drafts from platform data, 3 to 4 h per page (~70 h); Austin or a tuner fact-checks each, 30 min per page; Kamil templates and images 2 days; Kevin game-page routing and benchmark data pull 1 day | Optional: hardware-literate copy editor, 10 h a month at $30 to $60 an hour | ~110 |
| 3. Earn | Weeks 4 to 12, ongoing | Follow-ups on press requests; roundup pitches; first data story; Reddit and Discord presence | Vinson pitches and data story 6 h a month; Austin's team on community | Link outreach freelancer, 20 h a month at $25 to $50 an hour, written brief: no paid links, no networks, every placement reported | 20 to 30 a month |
| 4. Measure | Monthly from week 4 | One-page report; ten AI prompts; adjust the build list | Vinson 2 h a month | None | 2 a month |
Cost, outsourced portion: $500 to $1,000 a month for outreach from week 4; $300 to $600 a month for the editor if used; tools $0 (GSC, GA4, Bing WMT, Ahrefs WMT and Screaming Frog free tier all cover this stage). Total $500 to $1,600 a month against Bobby's $3,000 plus tools, and the in-house version produces the content, which his plan bills for separately.
Team time it does consume: about 90 hours of mine across the 90 days, front-loaded in weeks 2 to 8, roughly two days of Kevin, three of Kamil, and two to three hours a week of Austin or a tuner reviewing technical claims. The fact-checking is the one thing that cannot be skipped or outsourced: every performance and anti-cheat statement on these pages will be quoted back at us.
| Measure | Today | Day 90 |
|---|---|---|
| Non-branded search clicks a month (GSC) | 254 | 800 |
| Indexable pages with real content | 13 | 40 |
| Referring domains | 8 (Fluenta, 14 July; true figure after press is higher, to confirm in Ahrefs WMT) | +25 from press requests, +10 from directories |
| Search-first-touch new subscribers a month (platform) | 531 | 700, with non-branded share visible |
| AI category recall (ten-prompt test) | 0 of 5 engines | Named by at least 1 of 5 for "Hypertune vs" and safety prompts; category prompts realistically 6 months out |
| Press pieces linking | 17 of 108 | 40 of 108 |
Non-branded rankings for the big category terms ("game booster", "fps booster") are a 6 to 12 month outcome against sites with domain ratings of 65 to 91. The 90-day plan builds the pages and links that make that possible; it should not be judged on those terms at day 90.
Bobby (V1.4). Correct framework, generic execution, and it mis-states the site: proposes schema, FAQ answers and a guides section that already exist; says FAQ schema will produce rich results (withdrawn 2023); ignores the press run, Bing, and the existing /game/valorant template; includes creator, event and sponsorship scope that belongs to other budgets; has no targets. The $1,000 retainer has produced the Search Console baseline and the indexing count, both useful and both now reproducible in-house. If he stays, scope him to outreach only (section 6, item 3 to 5) with a monthly referring-domain target and a no-paid-links rule, at the current retainer. The $3,000 ask is for work this document does in house.
Fluenta. An automated lead magnet with one accurate conclusion (technically ready, not cited, fix is authority and content) and several errors: backlink data dated 14 July, the only "spam" domain is a legitimate press pickup, the disavow file would remove a real link, the suggested title tag is "game booster | Hypertune", and the FAQ copy is template text. Use its probe result (0 of 5) and its keyword volumes as inputs; do not act on its recommendations and do not upload its file.
Press coverage backlink check 2026-09-21.csv: 108 URLs, HTTP status, link status, action.