Searching

I asked ChatGPT to write a blog post about search.

Here’s the generated text for your perusal:

How to Search in May 2025 (and Actually Learn Something)

Search in 2025 is no longer about keywords and blue links. It’s about working with AI agents that think through your query, read the web, and return something far more useful than a list of URLs.

Whether you’re researching for work, writing content, or just trying to understand a complex topic, here’s how search works best right now—from my perspective as a long-time blogger and web explorer.

1. ChatGPT + o3 + Web Search

This combo is my go-to.

• Why it works: The o3 model is currently the most capable general-purpose AI OpenAI offers. When combined with search access, it behaves like a mini analyst—summarizing, clustering, and even generating content with a point of view.
• Great for: Thought leadership posts, growth strategy, market synthesis.
• Example prompt to try:

“Act as a B2B SaaS growth strategist. Think step-by-step: 1) grab the 10 best articles on buyer psychology 2) cluster their learnings into three themes; 3) craft one copy angle per theme that sells a $49 HR upskilling course. After thinking, output a single 70-word post that blends the strongest angle with a CTA—no emojis, no fluff.”

It’s like giving an intern a research brief—but they return in 30 seconds with something halfway ready to publish.

2. Grok + DeeperSearch (on X)

Yes, Grok—the AI chatbot integrated into X (formerly Twitter)—has quietly become a powerful search tool for real-time trends.

• Why it works: “DeeperSearch” lets you scan the firehose of breaking news, creator commentary, and policy moves, especially around AI.
• Great for: Headline scouting, social listening, quick post drafts.
• Example prompt:

“Example Q: ‘Top AI-policy headline today?’ Example A: ‘? EU weighs export caps on frontier models.’ Now—using DeeperSearch—write exactly 30 fresh X posts (?130 chars each, start with ?) on the biggest AI-founder moves announced today.”

It’s fast, punchy, and tailored to the social-media pulse.

3. Perplexity + Research Mode

Perplexity was once the “AI of search.” It’s evolved into a lean research assistant for professionals.

• Why it works: You can choose the model, cite sources inline, and generate multi-step reasoning—especially strong in B2B or startup contexts.
• Great for: Metrics, funding analysis, decision support.
• Example prompt:

“Begin as a skeptical angel investor reviewing 2025 funnel-conversion data on $7 tripwire offers; list three red-flag metrics with live sources. Then switch voice to a CRO-obsessed growth lead; for every red flag propose one AI-driven fix projected to lift CVR ?20%. Footnote every data point with link references.”

It’s like giving GPT a Harvard Business Review hat.

4. ChatGPT + Deep Research Mode

When you need to go deep—not fast.

• Why it works: “Deep Research” lets GPT generate long-form, structured reports, pulling from live sources when enabled.
• Great for: TAM analysis, diligence reports, technical overviews.
• Example prompt:

“Act as a private-equity analyst. Produce a 15-page diligence report on the total addressable market, growth drivers, competitive moats, and acquisition multiples for AI-powered LinkedIn-content tools aimed at SMBs. Use APA citations and close with a one-page investment memo summarizing risk-reward in plain English.”

This is real research, not just regurgitated blog summaries.

---

Final Thoughts

In May 2025, search isn’t just search anymore—it’s promptcraft. Knowing how to ask matters just as much as where you ask. And the best queries don’t just request information; they give structure, tone, and context.

— End of AI-generated text

I think one thing is clear. We are reading a ton of AI-generated articles in our daily streams.

Be wary.

Use AI like a productivity tool.

What won’t go away are:

ideas

opinions

insight

the need to create content for your company that describe your service or product in plain language.

the skill to position your service or product.

the need to generate citations and cross-references from 3rd party sites, e.g. press articles, podcasts, Reddit threads, 3rd party events and press coverage.

It seems only 1 out of 4 of the references AI uses are from a brand’s own site.