Your buyers research inside ChatGPT, Claude, and Perplexity now — and an agent can only recommend what it can actually parse. So I pointed an agent-readiness scanner at 106 semiconductor, analog, power, RF, sensor, MCU, memory, passive, connector, EDA, distributor, and test-&-measurement sites and scored each 0–100 across 20 signals (robots.txt, llms.txt, MCP, structured data, content negotiation).
The result: an industry-wide F
Average score: 42/100. Only one site cleared a B; zero scored an A.- A typical homepage costs an agent 40,000–90,000 tokens to read. The same content as clean markdown is ~1,000–2,500 — ~95% of it is wasted on markup an agent has to chew through.
| Most ready | Least ready | ||
|---|---|---|---|
| Quectel | 71 (B) | Cadence | 16 (F) |
| onsemi / Nuvoton / Silergy | 69 (C) | Sensata | 16 (F) |
| MediaTek | 63 (C) | Anritsu | 16 (F) |
| AMD / Infineon | 60 (C) | Goodix | 16 (F) |
(Texas Instruments 39, Analog Devices 34, NVIDIA 55, Qualcomm 39.)
The three failures almost everyone shares
1. No markdown content negotiation. Send Accept: text/markdown and you still get a wall of HTML.
if ((req.headers.get('accept') || '').includes('text/markdown')) {
return new Response(markdown, {
headers: { 'Content-Type': 'text/markdown; charset=utf-8' },
})
}
2. No llms.txt. The simple index that tells an agent which pages matter:
# Your Company
> One-line description.
## Products
- [CT8000 3D Hall sensor](https://you.com/ct8000.md): ±40 mT, I²C / SPI
3. No MCP server. None of the 106 expose an endpoint an agent can call (search_parts, get_datasheet), so it's stuck scraping rendered HTML.
Check your own
Paste any domain into the free scanner — same 0–100 score, the category breakdown, and the exact fixes, in ~2s, no signup: https://aireadify.ai/scan
Method/caveats: public pages only; bot-blocking sites excluded (so reality may be a touch worse); scores are a snapshot. Disclosure: we built the scanner and do agent-readiness work.