Used to open twenty tabs and hope an answer surfaced. Got burned enough times that I started building a source map before I wrote a word. Fifteen minutes upfront, saves me an hour later, and nobody emails me about a busted citation.
One warning before you run any of these: the models will invent sources. They’ll hand you a URL that 404s and quote an article that never existed. Verify before you cite anything. The prompts below are starting points, not finished research.
[Claude]
I am researching the 2026 shift in local SEO ranking factors. Given my question, list the 12 most authoritative sources (industry studies, Google statements, reputable blogs) that would settle it. For each, give the domain, why it is authoritative, and one specific claim it supports. Output as a ranked table with a confidence score per source.
[ChatGPT]
I need to verify a claim about AI adoption in small businesses before I publish. Given the claim, find three independent, primary sources that confirm or refute it. For each, quote the exact passage, note the publication date, and flag any bias. Output as a verdict summary: confirmed, disputed, or unverified, with your reasoning.
[Cursor]
I am building a research notes file for a blog post on remote work trends. Given my outline, generate a markdown template with sections for source URL, key quote, date, and relevance. Then write a script that scans my open browser tabs, extracts titles and URLs, and fills the template. Output the template and the script.
Full archive at aitoolalliance.com/tag/prompt-pack.