Case study · Client #1: ourselves

Copute was invisible in AI search. Here's the fix — with receipts.

Scan date: 2026-07-04 · Engine: Google Gemini (live web search) · Fixes shipped: same day · Re-scored: weekly

We sell AI visibility, so we ran the treatment on our own brand first. This page is the exact deliverable a COPUGEO Operator client receives: the before-state with evidence, every fix that shipped, and the loop that proves whether it worked. Nothing here is a mockup — every claim links to something you can check yourself.

Before · the scan

We asked Gemini the questions Copute's buyers ask

COPUGEO's scan battery asks an AI engine real buyer-intent questions — unaided, meaning Copute is never named in the question. Appearing in those answers is what AI visibility means. 7 of 13 questions completed before the engine rate-limited the scan; the rest re-run on the next weekly pass.

0 / 7
answers that mentioned Copute (unaided)
0 / 90
sources cited by Gemini that were copute.ai
0 / 100
Visibility Score at baseline*
Buyer question asked (verbatim)Copute in answer?Sources citedcopute.ai cited?
What is the best AI agent execution platform in 2026?✗ absent19
Recommend a AI agent execution platform. Give me your top 3 with reasons.✗ absent14
Top AI agent execution platform tools right now — which should I actually use?✗ absent13
Best AI agent execution platform for crypto and web3 founders?✗ absent18
I need a AI agent execution platform. What are my options and which is most trustworthy?✗ absent13
Which AI agent execution platform is worth paying for in 2026?✗ absent0
What are the best alternatives to Virtuals Protocol?✗ absent13
6 further questions (brand + comparison) hit the engine's rate limit mid-scan — queued for the next weekly run.
Q: "Best AI agent execution platform for crypto and web3 founders?" — Copute's exact target buyer
"For crypto and web3 founders seeking the best AI agent execution platforms, a growing ecosystem of tools and frameworks is emerging to facilitate auto…" — Gemini named the ecosystem and cited 18 sources.
Copute was not one of them. Neither was copute.ai among the 18 sources.
Gemini answered 7 questions about Copute's own category, citing 90 sources along the way — and not one answer or source was Copute. Every one of those answers sent a warm buyer somewhere else.

Why: the site gave engines nothing to read

What an AI crawler saw on copute.ai (2026-07-04, before)
Total machine-readable text on the page: the two words "copute.ai" (a 1.2 KB JavaScript app shell)
Page title: "copute.ai" · Meta description: none · Structured data: none
llms.txt: missing · robots.txt: missing · sitemap.xml: missing
Citable content pages an engine could quote: zero

An AI engine can only recommend what it can read. Copute's site — like most JS-rendered sites — was unreadable.

The fix · shipped same day

What the COPUGEO agent shipped (the Operator deliverable)

1 · Canonical description, synced everywhere

One authoritative description of Copute — "Community Compute for real AI execution" — now appears identically in the page metadata, structured data, llms.txt and about page, so every engine reads the same story.

2 · Full metadata layer on copute.ai

Real title, meta description and social-preview tags replaced the empty shell.

verify: view-source copute.ai →
3 · Structured data (JSON-LD)

Organization schema plus all four products (COPUGEO, Swarming, CoAI, Miru) as machine-readable entities — the format engines parse natively.

4 · llms.txt on every property

The emerging standard file that tells AI engines what a site is, in their language.

copute.ai/llms.txt · copugeo/llms.txt · swarming/llms.txt
5 · A citable content page

Engines cite pages with quotable text — a JS app gives them nothing. A static about page with the company story, product portfolio and FAQ schema now exists to be quoted.

copute.ai/about.html →
6 · robots.txt + sitemap.xml

Crawl directives welcoming AI crawlers, plus a sitemap pointing at the citable pages.

7 · Product subdomains optimized for their own categories

copugeo.copute.ai rebuilt as a crawlable landing targeting "generative engine optimization"; swarming.copute.ai given description, schema and llms.txt positioning it as a decentralised AI agent swarm.

8 · Crawlable homepage body

Most AI crawlers don't execute JavaScript — so the app shell now carries static content (description, products, links) that crawlers read and the app replaces on load. The homepage is no longer a blank page to a non-rendering crawler.

9 · Comparison & category content pages

Unaided queries are won by whoever wrote the clearest page. Shipped: an AI-agent-platform landscape page on copute.ai targeting the exact questions from the baseline scan, plus GEO-category comparison pages on copugeo.copute.ai.

copute.ai/best-ai-agent-platforms → · copugeo/compare · copugeo/best-geo-tools
10 · Crawl acceleration

IndexNow key hosted and pings submitted so engines that support it (Bing — which feeds ChatGPT — and others) are told to re-crawl now instead of whenever they get around to it. Google/Bing webmaster sitemap submissions queued.

After · verifiable now

Site state: before → after

DimensionBefore (2026-07-04 AM)After (2026-07-04 PM)
Machine-readable text on copute.ai"copute.ai" (2 words)Full description, product portfolio, FAQ
Meta descriptionnonecanonical Copute description
Structured data entities0Organization + 4 products + FAQ schema
llms.txt across Copute properties0 of 33 of 3
Citable pages engines can quote0copute.ai/about.html + copugeo landing
robots.txt / sitemap.xmlmissinglive
What we will NOT claim yet: a higher score. AI engines re-crawl and re-index on their own schedule — typically one to three weeks. The Visibility Score only moves when the engines' actual answers change, and that is the only receipt that counts. This page gets updated with the weekly re-scan deltas as they land.

The receipt loop

1 · ScanWeekly, every Monday: the same buyer questions, re-asked.
2 · ScoreVisibility /100 = 50% unaided mentions + 30% citations + 20% brand accuracy.
3 · DeltaWeek-over-week change, per question: which answers were won, which were lost.
4 · ReceiptEvery shipped fix is judged by the next scan — no unverifiable claims, ever.
WeekVisibility ScoreUnaided mentionscopute.ai citedNotes
2026-07-04 (baseline)0 / 100*0 / 70 / 90 sourcesfixes shipped same day
next weekly scanappended automatically — watch this space

* Baseline computed from the 7 completed answers; 6 rate-limited questions re-run next scan. A partial baseline understates nothing here: zero mentions is zero mentions.

Your brand is probably in the same "before" column.

Find out in 60 seconds — free scan, no signup. Then decide if you want the same treatment.

Run your free AI visibility scan →

Monitor $299/mo · Operator $999/mo (founding prices) · rencelin@gmail.com