Two weeks of AEO: what we did, and why

1. Choosing the questions Som researched · Vin approved · Hoa loads

1. Provisional 25 questions approved on 20 Oct, so tracking could start the same day instead of waiting for research.
2. Som's research on 25 Oct: demand and buyer intent from Semrush, Peec, Brand Radar and Search Console, with five proposed swaps and one line of evidence each.
3. Vin approved four swaps on 27 Oct. In: best game booster for PC · improve FPS without upgrading hardware · safest PC tweaks for better FPS · Hypertune vs EXM Tweaks. Hypertune vs Razer Cortex stays.
4. The set is now fixed until 31 Dec.

The swaps are not loaded into Brand Radar, Peec or Otterly yet (Hoa, E1-8). Run 2 waits for it.

2. One set of facts Som wrote · Hoa, Vin and Travis check · Austin signs off

1. Facts page drafted on 22 Oct from the approved facts: price, offers, cancel, install, partners, figures and their tests.
2. Handover notes on 26 Oct: 40 facts, each with its source and how sure we are, so the checkers know what to verify.
3. Decided on 28 Oct that it stays internal. The trust page gets the partner and uninstall lines (Travis). /pricing already has price, refund and cancel.
4. Old fact sheet retired on 28 Oct, and its directory listing copy moved onto the facts page, so there is one document instead of two that drift.

Hoa's check (E3-50, due 27 Oct), then Vin and Travis, call 30 Oct, Som's round 2 on 2 Nov, Austin signs off 4 Nov.

3. Tracing wrong answers Hoa flags · Som traces

1. Hoa flagged the AMD miss: every engine leaves out that we are an AMD partner.
2. Som traced it on 28 Oct: no public page says "AMD partner". Launch press (GamesBeat, Tom's Hardware, TechPowerUp, TweakTown, MCV) says Intel partner and AMD supported. Our site only says "official Intel and AMD toolkits".
3. Two more sources logged: the homepage line "never touching game files", and TechPowerUp's old yearly price of $59.99.
4. Fix agreed in principle: one sentence on the homepage, /technology and /what-hypertune-changes saying Intel partner and AMD partner, supports NVIDIA.

Austin confirms we can say "AMD partner" publicly. Run 2 may flag more.

4. Pages AI can quote Som set the rules · Som and Hoa write

1. Nine answer rules approved 19 Oct, written into the AI content framework and the writing skill on 22 Oct, and sent to Travis on 23 Oct so his reviews use the same standard.
2. Word ranges per page type and the "stress test" wording added. The app's test is now an optional before-and-after stress test with no number shown, so no page calls it a benchmark.
3. A fair free-tools table in the Best PC optimizer draft (NVIDIA App, AMD Adrenalin, MSI Afterburner), because AI often recommends a free tool instead of us.
4. A test method sheet drafted, so any FPS figure can show how it was measured. Press and AI do not trust a number without one.
5. Eleven drafts written to the rules: CS2, Valorant, Fortnite, Warzone, gaming guide, GPU scheduling, buyer page, How to increase FPS, Best PC optimizer, Apex, Rust.

Austin's FPS test evidence (owed since 22 Oct). Homepage and /pricing opening lines with Vin. Hoa confirms she is writing to the nine rules.

5. How every page gets checked Som and Hoa built it · Bobby, Vin, Travis, Austin use it

1. One order for every article: writer, then the other writer, then Bobby's SEO check, Vin's final run, Travis's written feedback, a 30-minute call, round 2, Travis's re-check, and Austin's sign-off. Set 20 Oct and put into the calendar on 22 Oct.
2. One combined checking prompt, shared on 25 Oct, so Som and Hoa check the same way and their notes look alike.
3. Every handover lists the primary keyword, secondary keywords and the H2s and H3s they target, so Bobby's SEO check happens before Travis and not after.
4. The checker sends an edit list as a document, and only the person who owns the draft link publishes. That way two people never overwrite the same page.
5. Used on all eleven drafts in the two weeks.

FPS and Apex were ticked by their own writer because the other read had not come in. Hoa reads both before Bobby. Edit lists only work if the document is shared, which is decision 4 below.

6. Testing an AI writer and a research tool Som

1. Semrush trial used for question research: what people ask AI and how it describes competitors. That evidence chose the four question swaps.
2. Semrush exported (14 topics, 73 cited sources), then cancelled on 25 Oct before a $495.72 charge.
3. Profound's AI writer drafted the buyer page from the same brief as Som's. The buyer page was picked because Som's draft was finished the day the trial began, so the comparison was fair.
4. Result: 0 fact errors and a better "worth it" section, but four lines break our rules, "up to 60%" six times, and no date, Terms link or cancel cutoff. Our draft stays the base.

Decision on the call: do not pay, and take two lines into the buyer page.

7. Reporting what AEO achieves Som writes · Vin reviews · Kevin builds the tracking

1. The report. One report every two weeks for Austin and Jerad, covering SEO and AEO. Report 2 is due 30 Oct. Som writes the AEO part.
2. The tracking. A brief for Kevin, who runs our data. It asks him to remember which AI engine brought a visitor, carry that through signup and payment, and match it with what the customer says in the signup survey. Then each paying customer from AI is counted once.

1. AEO part of Report 2 drafted on 28 Oct: our numbers against the targets (named on 13 of 25, 34 AI visits, 15 "AI chat" signups), what moved, the AMD wrong answer, and next actions with owners.
2. Tracking brief for Kevin written on 28 Oct. It counts a paying customer from AI three ways, then once:
   1. Last visit: the AI engine on the visit where they paid.
   2. First visit: the first AI engine that ever brought them, remembered for 90 days.
   3. What they say: their answer to "How did you hear about us?", with "AI chat" split into ChatGPT, Perplexity, Google AI answer and other AI.
   4. One count: a customer counts once if any of the three says AI.
   It also keeps AI crawler hits apart from real visitors. It notes that clicks from Google's AI Overviews arrive as normal Google traffic, so the survey is the only way to count them.
3. Vin mocked up an AI tab for the dashboard on 20 Oct, and Austin passed it to Kevin. That way the AI numbers show where the team already looks, not only in the report.

Som sends the AEO part to Vin before Report 2 on 30 Oct, and the tracking brief to Austin for Kevin. Kevin confirms how trials and refunds count.

8. Measuring with Brand Radar Som built · Hoa logs · $100 a month · kept

1. Report "Hypertune AEO" built on 22 Oct: 25 questions weekly on six engines, plus 5 buying searches daily, against Hone, Razer Cortex, Radiant, EXM Tweaks, Paragon Tweaks, Process Lasso and ExitLag.
2. Growth package bought, because our plan's 150 free checks a month were too few for about 720 a month. ChatGPT index bought for the wider view.
3. Weekly export agreed instead of an API pull, so the data arrives the same way each week.
4. Run 1 logged: all 25 questions. Partner question read on 27 Oct, which is how we found the AMD miss.
5. Signed-out spot checks in ChatGPT and Perplexity, to confirm the tool matches what a real buyer sees. The safety question hit a sign-up wall.

Run 1 still needs the partner cell, the Perplexity row and positions, or the next run has nothing to compare with. Then load the swaps and log run 2. The AI Overview study brief is due 30 Oct: what the pages Google's AI cites have in common.

9. Otterly Hoa · $38 a month · test to 19 Nov

1. Bought on 20 Oct with the AI Mode add-on, because Google AI Mode is one of our six engines. Moved to Austin's card on 22 Oct.
2. 15 of our 25 questions loaded into its prompt tracking, as a cross-check on Brand Radar.
3. Two drafts scored in the content checker. Gaming guide, 21 Oct: static 100%, content 75%, structured data 53%. Valorant, 24 Oct: static 100%, content 74%, structured data 84%. These were the first two drafts the nine rules were applied to, so the scores show whether the rules work before we write more pages the same way.
4. Renewal date, 20 Nov, written in the tool sheet so it cannot renew without a decision.

Only 2 of the 1,000 audits this month are used. Turn the Valorant fixes into an edit list Som can apply, or the score changes nothing. Then score every AEO draft before Bobby's check and again once live, until 18 Nov. The gaming guide's structured data, at 53%, is the weakest score so far.

10. Peec and Scrunch Vin set up · Hoa runs Peec · Vin runs Scrunch · test to 19 Nov

1. Both bought on 20 Oct and set up: Peec with our topics, brand profile and US market. Scrunch with description, personas, five competitors and topics.
2. Setup check on 25 Oct found Peec missing EXM Tweaks and Paragon Tweaks, and fixed it.
3. Both moved to Austin's card on 22 Oct so tracking did not stop when the trials ended.
4. Peec's question volumes were used in Som's question research.
5. Peec's Perception page already shows how AI describes us: strongest on PC optimization (94) and competitive gaming (89), weakest on anti-cheat safety (12), the attribute where Peec says we compete best.
6. Hoa read Scrunch on 24 Oct: Hypertune is mentioned in 51.6% of answers across all prompts, but 0% on non-branded prompts. Its audit of /technology scored 100 and missed the line "nothing inside your games", so a perfect score does not mean the page is right.

Peec: the weekly timed side-by-side with Hoa's manual fact check starts now, and the swaps need loading. Billing dates for both go in the tool sheet. Scrunch: one page rewrite and the audit findings by 18 Nov.

11. Getting named on other sites Hoa · Som · Bobby

1. Listing correction brief sent to Bobby on 26 Oct: Capterra, G2 and Trustpilot. Hypertune.gg gets its own profile, separate from hypertune.com, which Capterra was describing.
2. YouTube list for Travis: our five most-viewed videos, with titles, chapters and captions to fix, because YouTube is one of the most-cited sources.
3. 73 cited sources exported from Semrush, as the starting list for the citation map.
4. Peec has already proposed 29 places to get mentioned, such as Tier1Settings' safe PC optimization rankings and a Digital Trends gaming guide. None are actioned yet. They feed the citation map.

Citation map v1 (Hoa, due 29 Oct): every site AI cites for our questions, so outreach starts where it moves answers. The YouTube list goes to Austin for Travis. Wikipedia sits with Bobby on the SEO tracker (W1).