RajanChavada

Computa

[RajanChavada @RajanChavada]

@imryanarya007 https://t.co/zwPt3WkoiD

[RajanChavada @RajanChavada]

@spacedevin https://t.co/zwPt3WkoiD

[RajanChavada @RajanChavada]

Checked 1,284 homepages from 10 seed funds' public portfolios.

66% have no Organization schema: nothing machine-readable that says the company's name, logo and official profiles.

Every fund was between 54% and 76%.

https://t.co/AmGvV8xiVv

[RajanChavada @RajanChavada]

Asked ChatGPT and Perplexity "what does [startup] do?" for 240 YC companies.

Homepage with almost no text unless JavaScript runs:
Perplexity cited the site 76% of the time
ChatGPT 98%

Readable homepage: both ~100%.

https://t.co/b447KpXl1L

[RajanChavada @RajanChavada]

I built an agent to fix robots.txt files that block ChatGPT.

Then I checked 4,336 YC startups. 16 of 4,000 block it.

I'd built a fix for a 0.4% problem. What I found instead: https://t.co/4ZYE4tYDTB

[RajanChavada @RajanChavada]

I checked every active YC startup's homepage the way an AI crawler reads it: no JavaScript.

1 in 8 have almost no text. To ChatGPT, Claude and Perplexity they're close to empty pages.

4,336 sites, full data: https://t.co/Fy1ms5Uk9q

[RajanChavada @RajanChavada]

@ClaudesRH Love it

[Tenuo @TenuoAI]

The prompt injection worked.

The unauthorized action didn’t.

Rajan built an AEO agent with Temporal + Tenuo that can inspect a repo, make changes and open a PR.

In the demo, it takes the bait and tries to read .env.

Tenuo denies the tool calls. The agent carries on with the job.

This is the kind of agent we want to make possible: one you can actually trust to take action.

https://t.co/EWRx7JBCKn

[Tenuo @TenuoAI]

First @DecagonAI launched PACT.

Then @AIatMeta + @SierraPlatform announced PAP with @Genesys , @Shopify , @stripe , @Walmarttech and others.

Both help a business answer: “Did this customer authorize this agent?”

The next question is: “Authorize it to do exactly what?”

If I say “rebook Friday, aisle seat, ask me above $300,” those terms should travel with the agent and be enforceable by the airline.

That’s where we think Tenuo fits.

PACT/PAP get you to the customer’s yes. Tenuo carries what the yes covered.

[RajanChavada @RajanChavada]

@harleyfoote_ @TenuoAI Leaks: warrants are bound to the agent (PoP), so a stolen one is useless. To kill one mid-run, Tenuo Cloud has a signed revocation list.

Wrong-but-valid scope: the model can’t request scope. Each phase gets a minted warrant that can only narrow.

Docs: https://t.co/ypePly5PCA

[RajanChavada @RajanChavada]

Building an AEO agent that edits client sites and opens PRs. It reads untrusted pages and holds a GitHub token.

Prompts can't stop injection, so I added @TenuoAI. Every tool call needs a scoped warrant. The model never holds a key.

Try to break it:
https://t.co/LVHfpX9ARo

[RajanChavada @RajanChavada]

Tip for your top 5 pages: put an update date on your calendar.

AirOps found pages left 3+ months without updates are 3x likelier to lose AI visibility.

Refresh one page this weekend: new numbers, new examples, a dated changelog line.

#SEO #AEO

[RajanChavada @RajanChavada]

A thin Trustpilot profile does a lot in AI search.

Seer looked at 804K AI answers: brands with just 1 to 13 reviews went from a 1% to a 53.5% citation rate.

Cheapest AI search fix most brands skip.

#AISearch

[RajanChavada @RajanChavada]

Has a customer ever told you ChatGPT sent them?

Trying to figure out how many businesses actually see this vs just read about it.

[RajanChavada @RajanChavada]

Who still uses bing these days

[RajanChavada @RajanChavada]

Free AI visibility check, 10 minutes:

1. Write 10 questions your buyers would ask ChatGPT
2. Ask each one 3 times
3. Count how often you're named
4. Open the links it cites when it names a competitor

Those links are your outreach list.

[RajanChavada @RajanChavada]

Schema won't get you into ChatGPT on its own.

Ahrefs matched 1,885 pages with schema against 4,000 without. No significant citation lift in ChatGPT or AI Mode.

Reviews and mentions on other sites move it more.

[RajanChavada @RajanChavada]

See where your brand stands across 4 AI assistants, free: https://t.co/HxSKInTb9L

[RajanChavada @RajanChavada]

We tracked 6 small brands across 801 ChatGPT answers.

When ChatGPT named them, it almost always recommended them. "Mentioned but not recommended" was 0 to 2%.

The real problem: they were missing from 54% to 89% of answers.

#SEO #GEO

[RajanChavada @RajanChavada]

Showing up in ChatGPT is a coin flip you can't see.

SparkToro found the odds of getting the same brand list twice are under 1 in 100.

Track a % across many runs. A single screenshot proves nothing.

#AISearch