Slide 1 · Cover (full-bleed brutalist structure · canonical messaging)
From prompt to proof.
The brand layer for the agent web.
Slide 2 · Market shift (one stat object)
The shift
Agents now read the web.
Sample publisher audit
63%
of crawl traffic tagged agent-driven — and in this sample model the existing ad stack does not monetize it.
Slide 3 · Mechanism (clean blocks · served-at-edge emphasised)
How it works
One request, from page to payout.
01
Request
named agent user-agent
→
02
Fragment match
placement + rules
→
03
Card served · edge
match or no-fill · target <50ms
→
04
Value + proof
payout · signed · logged
→
05
Reporting
per agent · fragment
Slide 4 · Section (full-bleed circulation · operational caption only)
From request to receipt
Every request moves through structure, and leaves signed proof.
Slide 5 · Reporting (chart reconciles with table · sample labels)
What publishers see
Demand, by agent — sample 7-day window.
| Agent | Requests | Fill |
| OpenAI | 6,204 | 78% |
| Anthropic | 4,118 | 74% |
| Google | 3,902 | 69% |
| Perplexity | 2,560 | 61% |
| Meta | 1,708 | 52% |
OpenAI + Anthropic = 56% of sample agent requests · weighted fill ~70%.
Slide 6 · Close (concrete next action)
Next
Run a pilot
this quarter.
One site, one fragment, first agent reporting in ~a week. From prompt to proof.
Slide 7 · Appendix (sources & credits — the visuals + the claim sources)
Appendix
Sources & credits.
Imagery · CC0 / public domain
Brutalist facade — Osman Rana · CC0, Wikimedia Commons
Muziekgebouw aan 't IJ — Fons Heijnsbroek · CC0, Wikimedia Commons
Oosterdokseiland columns — Fons Heijnsbroek · CC0, Wikimedia Commons
Indigo duotone applied across all imagery; full register in design/assets/images/image-register.md.
Claims & data
All metrics in this deck are sample / illustrative — not production results.
AI crawlers & agent user-agents are documented: developers.openai.com/api/docs/bots · Google-Extended · developers.cloudflare.com/bots.
robots.txt is a protocol, not access authorization — RFC 9309.