For AI safety and research organisations
Your site is how a policy adviser, a journalist and your next hire decide whether you are serious.
A program officer at Coefficient Giving does their own research before a grant, and the site is where it starts. A policy adviser at the AI Security Institute opens it before a meeting. A journalist opens it the week an incident makes your field news. A candidate opens it against three other offers. An AI model reads it when someone asks who works on your problem.
None of them email first.
What most research sites give them
- A site a researcher built in a weekend, on a static generator, a Notion page or Squarespace, and has not touched since.
- Publications as a list of arXiv links with no abstracts, no authors on the page, and nothing a model can cite.
- A team page that stops at the founders.
- No Form 990, no accounts, no board, no funders named.
- A careers page that is a link to a Google Doc.
- A US entity and a UK entity that the site never explains.
Each of these is read as a signal. The signal is that the organisation is not yet built to be looked at.
What I build
- A publications system where every paper has its own page, structured data, authors, abstract and a citable URL. Answer engines quote it instead of a third party’s summary.
- A team page with roles, affiliations and what each person works on, maintained by the people themselves.
- Governance pages: entity structure, board, funders, 990 or accounts, conflict of interest policy, in the place a program officer looks.
- A careers section that a researcher can publish to without a developer.
- A news and events section that survives the person who set it up leaving.
- Astro, static output on Cloudflare, a CMS your team edits, all in your GitHub organisation from day one. No proprietary layer, no build that runs only on my machine.
How it works
The same two steps as every engagement. A Blueprint Audit reads the site the way each of those five readers does and scores ten areas. Then a fixed proposal, or a recommendation against one. Details on the process page.
For a research organisation the audit adds a US 501(c)(3) module and a UK module where both entities exist, and a publications review: what is findable, what is citable, and what an AI model currently says about you when asked.
Questions
Yes. Billing in USD, a contract under English law, and the jurisdiction and insurance position confirmed in writing before anything starts.
You may not need to. The audit says whether the problem is the platform or the content model. If your engineers can keep it, the roadmap tells them what to change and I do not need to be involved.
Enough to build for it. I work with FAR.AI, and I read the papers. I do not write your research pages. Your researchers do, in a CMS built so they can.
Two clicks. Three things.
Send the address of your website. Within two working days you get a short email naming the three things a funder would hold against it, including what AI assistants like ChatGPT and Google’s AI Overviews currently say about you. No report, no call, no follow-up unless you ask.
Received. Your three things arrive within two working days.
That did not send. Email hello@socialectric.com with your website address and I will reply from there.
If your site is the first thing a program officer, a policy adviser or a candidate reads about you, a Blueprint Audit reads it the way each of them does and tells you in writing what they will find.