AI prompting best practices for finding top talent

This blog summarizes key highlights from our recent webinar with Gem’s AI experts on AI prompting best practices in recruiting. Watch the full on-demand recording here to learn more about these strategies and see live examples.

When we polled our 400+ live webinar attendees about their AI recruiting experience, the results were telling: 60% are already using AI but want to level up their game. Most recruiting teams are trying AI, yet are getting mediocre results. The difference between "meh" AI outputs and game-changing productivity gains is that context is king.

Why traditional Boolean searches fall short

Traditional Boolean search relies on exact keyword matches. If you search for "software engineer OR programmer OR developer AND full stack AND startup," you'll only find candidates with those literal words on their profiles. This means you miss great candidates who don't use those specific terms, or worse, you find candidates who stuffed their profiles with buzzwords.

Modern AI tools, powered by large language models (LLMs) such as OpenAI, Claude, or Gemini, are helping recruiters find better candidates by going beyond keyword searches. They’re designed to understand semantic meaning, so when you prompt: "Find me software engineers with three years of full-stack experience at early-stage tech startups," the AI immediately understands what constitutes an "early-stage tech startup" and matches candidates based on that context.

While LLMs are powerful, they’re also generalists and rule followers. Therefore, mastering the art of prompting has never been more important.

AI prompting best practices

Think of AI prompting like giving detailed instructions to a new teammate. Just as you wouldn't tell a new recruiter to "find some good product managers," you shouldn't give vague instructions to AI.

Here’s a 3-part masterclass on how to level up your AI prompts: 

AI prompting 101: Getting started right

Be specific, not vague

• Instead of: "Write a message to a software engineer" 

• Try: "Write a message to a staff-level engineer who just got promoted. They value growth opportunities and are interested in scaling challenges."
  

• Instead of: "Find senior backend engineers"

• Try: "Find senior backend engineers with infrastructure experience who've worked at B2B SaaS companies and have experience scaling systems to handle high traffic."

Add context like you would for a new teammate

Include details about:

• Company stage and culture fit

• The industry or customer type they should have served

• Specific tools, skills, or technologies

• Mission alignment or values match

Example: "We're hiring for a Series C startup building AI infrastructure. Looking for senior engineers familiar with distributed systems who thrive in fast-paced environments where they can have direct impact on product direction."

AI prompting 201: Leveling up your prompts

Iterate and refine your prompts

AI responds well to feedback. Treat it like a collaborative conversation:

• "Make the tone warmer and more conversational"

• "Focus more on the growth opportunity rather than compensation"

• "Make this shorter for a LinkedIn InMail"

Tell AI what to avoid

Be explicit about what you don't want:

• "Exclude candidates with only academic experience"

• "Don't use buzzwords like 'rockstar' or 'ninja'"

• "Avoid overly formal language"

Ask for multiple options

AI excels at generating variations. Request:

• "Give me three versions: one playful, one direct, one focused on mission"

• "List different candidate types who might fit this role even without the exact title"

AI prompting 301: Pro tips

Advanced technique: Two-shot prompting

Give AI two examples of ideal outputs to establish patterns. This works exceptionally well for:

• Candidate search: "Here are two ideal candidates for this role: [Profile A] and [Profile B]. Find more candidates with similar backgrounds and experience levels."

• Outreach messaging: "Here are two high-response messages I've sent to similar candidates: [Message 1] and [Message 2]. Generate a similar style message for a senior engineer joining our infrastructure team."

Real-world examples: Before and after prompts 

The difference between generic AI outputs and breakthrough results comes down to the context you provide. 

Einas shared practical examples of how TA teams transformed their AI workflows with prompting best practices. 

Personalized outreach

Generic prompt: "Write an email to a software engineer about a job opportunity."

Result: Generic, templated message that could be sent to anyone

Context-rich prompt: "Write a personalized outreach email for Sarah Chen, a senior software engineer at Airbnb with six years of experience building user-facing features. She recently posted about accessible design. The role is a Senior Frontend Engineer at our fintech startup (Acme), building tools for small businesses. Mention her specific background and connect it to our mission of democratizing financial tools." [Resume attached]"

Result: Personalized message that references the candidate’s specific interests and connects them to the company's mission

Pro tip: Upload a list of prospects and context about each to have AI create a personalized sequence for every candidate. 

Advanced sourcing

Basic prompt: "Help me find a good candidate from a top company for a machine learning role."

Result: One profile of a Machine Learning engineer, but not the right fit for the specific job.

Advanced prompt: "Find a senior ML engineer for an autonomous vehicle startup (Series B, 150 employees) building perception systems for self-driving trucks. The ideal candidate has experience with computer vision in production environments, sensor fusion, and comes from companies that have deployed ML in safety-critical applications."

Result: Highly targeted profiles that match specific technical requirements and industry experience

Pro tip: Give the AI detailed context about your company stage, product, and ideal candidate profile to get the most actionable sourcing results.

The ROI of getting AI right

Teams implementing these prompting best practices see significant results:

• 2x improvement in qualified matches within just 11 days of implementing better prompting practices

• 5-6x more qualified candidates from AI-powered sourcing compared to traditional methods

• 30-40% higher response rates from personalized AI-generated outreach

• 5x efficiency gains in application review when AI has proper context

Key takeaways:

1. Start with context: Give AI as much relevant information as you would give a new team member

2. Be specific: Replace vague requests with detailed, nuanced criteria

3. Iterate: Treat AI as a collaborative partner and refine your prompts based on results

4. Use examples: Show AI what good looks like with 2-3 concrete examples

5. Set boundaries: Explicitly state what to avoid or exclude

AI’s future in recruiting

The recruiting teams seeing the biggest wins aren't just using AI as an add-on feature. They're adopting AI-first platforms that integrate these best practices into every workflow from sourcing and application review to personalized outreach and interview scheduling.

The key is moving beyond fragmented point solutions toward integrated AI that understands complete candidate context across your entire recruiting process.

Ready to see these AI prompting best practices in action? Gem's AI-first recruiting platform has these techniques built into every workflow, helping teams boost recruiter productivity by up to 5x while cutting technology costs by 30-50%. 

Request a demo to see how leading organizations are scaling their recruiting with AI.