These numbers are measured from live production websites, not theoretical estimates. We crawled our own sites the way an AI agent would, then compared the processing cost to a single .well-known/ai manifest fetch.
What AI Processes Today — Per Query About One Business
| Page Crawled |
Size |
AI Tokens |
Cost* |
| Homepage | 62,622 bytes | 7,264 | $0.018 |
| Shop / Products listing | 64,836 bytes | 4,754 | $0.012 |
| About page | 26,478 bytes | 2,234 | $0.006 |
| Contact page | 34,837 bytes | 2,680 | $0.007 |
| 6 individual product pages | 236,184 bytes | 19,615 | $0.049 |
| Total without .well-known/ai |
424,957 bytes |
36,547 tokens |
$0.091 |
| With .well-known/ai (1 file) |
6,059 bytes |
540 tokens |
$0.001 |
*Cost based on GPT-4o input pricing ($2.50/million tokens). Claude and Gemini rates are similar.
At Scale — What This Means for AI Companies
| Daily Queries |
Without |
With |
Saved |
| 10,000 |
$912 |
$14 |
$898 |
| 100,000 |
$9,120 |
$135 |
$8,985 |
| 1,000,000 |
$91,368 |
$1,350 |
$90,018 |
| Annual (1M/day) |
$33.3M |
$493K |
$32.8M |
| Bandwidth Per Day (1M queries) |
| Without .well-known/ai |
395.8 GB |
| With .well-known/ai |
5.6 GB |
| Saved daily |
390.1 GB |
| Saved annually |
142.4 TB |
Speed Comparison
Without .well-known/ai: ~3.5 seconds
| DNS lookup | 50 ms |
| Crawl 10 pages | 2,000 ms |
| Parse HTML | 500 ms |
| Extract entities | 1,000 ms |
| Total | 3,550 ms |
With .well-known/ai: ~150 milliseconds
| DNS lookup | 50 ms |
| Fetch 1 JSON file | 100 ms |
| Parse JSON | 1 ms |
| | |
| Total | 151 ms |
Accuracy — From Guesswork to Guaranteed
| What AI Needs to Know |
Without (crawl + guess) |
With .well-known/ai |
| Company name & legal entity |
~85% — may confuse DBA names |
100% — explicit JSON field |
| Leadership & credentials |
~70% — may miss or fabricate |
100% — structured array |
| Products & pricing |
~60% — often outdated or wrong |
100% — typed, priced, linked |
| Trust & certifications |
~40% — cannot verify claims |
100% — cryptographic attestation |
| Overall company profile |
~50% accurate |
100% accurate |
Why AI Engines Will Adopt .well-known/ai
| Reason |
Detail |
| They already crawl it |
AI agents follow the same HTTP paths as search crawlers. /.well-known/ is an IETF standard (RFC 8615) — the same standard used for SSL certificates, Apple app associations, and security policies. Every crawler already knows this path. |
| It saves them money |
As shown above, structured data costs 98.5% less to process. At scale, that is tens of millions in annual compute savings. AI companies are actively looking for ways to reduce inference cost. |
| It makes them better |
Users are losing trust in AI answers about businesses because of inaccuracies. Structured, signed manifests eliminate guesswork and give AI engines a provably correct answer — improving their product. |
| Precedent exists |
Google adopted schema.org structured data. Bing adopted IndexNow. Every search engine adopted robots.txt and sitemaps. When a standard saves everyone money and improves quality, adoption is inevitable. |
| We submit for them |
Our automated pipeline submits every validated site to Bing, Yandex, and IndexNow-compatible engines weekly. AI engines that crawl these search indices automatically discover .well-known/ai manifests. |
The bottom line: For every 1 million AI queries about businesses, .well-known/ai eliminates 424 GB of unnecessary crawling, saves $90,000 in compute costs, delivers answers 23x faster, and guarantees 100% accuracy. AI companies will adopt this for the same reason they adopted every previous web standard — it makes their product better and cheaper at the same time.