Google Plays Down AEO as New $26M Web Index Targets AI Agents
Google said this week that generative AI answers require no special optimization playbook and confirmed its crawlers fetch JSON without parsing it, even as a newly funded startup built a 100-billion-document index meant specifically for AI agents. Separately, OpenRouter data shows agents already burn five times more tokens than human users per task, and Pew Research found AI-authorship signs on one in ten webpages.
This Week's Signals
Google Says No Special Work Needed for AI Search Answers
Why it matters for your score
Google's John Mueller said there is nothing special to do for generative AI responses in search, distinct from ordinary optimization. It is a reminder that Google still treats AI Overviews and AI Mode as extensions of classic ranking rather than a separate discipline, so businesses chasing AEO-specific shortcuts should keep crawlability, content quality, and structured data as the foundation instead of looking for a hack Google says does not exist.
Google Confirms Its Crawlers Fetch JSON but Never Parse It
Why it matters for your score
Gary Illyes clarified that Googlebot downloads JSON files, including JSON-LD structured data, but hands parsing to indexing systems later on shared infrastructure. Sites depending on schema markup for AI and search visibility should not assume the crawl itself reads or validates that markup; a malformed script can sit unnoticed until indexing catches it, or never does.
Startup Raises $26M to Build a Web Index Built for AI Agents
Why it matters for your score
Keenable's index already covers more than 100 billion documents and is used in production by AI labs and inference providers, built as infrastructure for agents rather than adapted from consumer search. As agent-facing indexes like this become a standard layer between a business and an AI answer, how content is structured for machine retrieval starts to matter alongside how it ranks on Google.
AI Agents Burn Five Times More Tokens Than Human Users, OpenRouter Data Shows
Why it matters for your score
Agents reload tool definitions, system prompts, and schema constraints on nearly every turn, which is why over 85% of their token use comes from repeated cached context rather than new output. Lean pages and clean, minimal schema matter here: agents are already an expensive audience to serve, and bloated or malformed markup adds cost an agent may simply skip past.
Pew Finds AI Authorship Signs on One in Ten Webpages
Why it matters for your score
Among pages published since ChatGPT's launch, the share showing detectable AI-authorship signs rises to over a third, concentrated on commercial .com domains. As AI-drafted content becomes the default state of the open web rather than the exception, differentiation shifts from publishing volume to verifiable specifics and clear sourcing that both readers and agents can point to.
Emerging Audit Signals
What to do this week
- 1
Re-validate your JSON-LD this week for syntax errors; Google's crawler fetches the file but does not check it, so mistakes surface later at indexing rather than at crawl time. Faro's /tools/ai-schema-creator can generate and check markup in one pass.
- 2
Audit page weight and script bloat on your highest-value commercial pages; if agents already pay a token premium just to read a page, unnecessarily heavy pages are the first thing an agent truncates or skips.
- 3
Assume a growing share of competitor content is AI-drafted, and differentiate with specific numbers, named sources, and claims a reader or an agent can actually verify against the original.
Sources
- ↗ Google: There Is Nothing Special To Do For Generative AI Responses In Search
- ↗ Google Says Google's Crawlers Don't Parse JSON
- ↗ Accel-backed Keenable is indexing the web for AI agents
- ↗ AI agents use five times more tokens than humans, OpenRouter data shows
- ↗ How much of the internet is written with AI?