How I use AI to support Public Relations

After narrowly losing yet another pitch earlier this year, I added up the time I had spent preparing proposals: 300 hours in four months. I wanted to know whether we were pursuing the right business.

Using Gemini and a script in Google Docs, I compared five years of proposals with our client invoicing records. One category of prospective client had produced no wins. We now decline that work, and the analysis informs a tool that scores each new brief before we start a proposal.

That decision changed where we spend our time. It also prompted me to develop how I use AI across the consultancy: assessing prospects, researching their businesses, developing recommendations and checking how our communications are understood.

I founded and run Pinpoint PR, a communications consultancy in Singapore. Over a thirty-year career, I have developed methods for understanding a client’s business and deciding what communications can contribute. This year, I have spent scores of hours teaching the AI tools I use how to apply those methods.

Research before recommendations

Before working on a proposal, I run a competitive intelligence skill file I have written. It gives the AI a sequence to follow, with a different specialist perspective at each stage.

The AI examines the company and its regional presence, analyses its media coverage, investigates competitors and identifies communications opportunities. I also ask it to consider the business from a revenue and sales perspective: who should its priority customers be, and how could it reach them? The assessment then maps the stakeholders influencing its success.

I also supply a coverage export from our media monitoring platform, and the AI gathers additional material. Within 20 to 30 minutes, I have an initial appraisal to investigate and challenge. Previously, assembling the research for that assessment could take weeks.

For a recent prospect, the analysis included more than 30,000 customer and employee reviews alongside a year of media coverage. The reviews presented a more mixed picture than the coverage suggested. That difference influenced the programme we proposed.

Experience and instinct still matter, but now I have more evidence with which to test our ideas.

Bring the team’s thinking into the work

The broad research gives my colleagues and me a basis for discussing the findings and developing recommendations.

I record and transcribe our conversations with an AI, then bring the transcripts into other AI models alongside information we received from the prospective client. That helps organise our thinking and outlines the research into a proposal we can examine.

Using AI for research and strategy is already common. Muck Rack’s survey of 564 PR professionals found that 76% use generative AI. Among reported uses, editing accounted for 86%, research and insights for 76%, writing for 74%, and strategy and planning for 68%. Muck Rack’s State of AI in PR 2026

My own investment has been in developing our own working process within the tools. A sequence of prompts builds on earlier findings, so the entire body of research informs the messaging, strategy and proposed activity. It means that we can spend more time examining the implications of our proposed strategy.

I also use AI to build a claims register alongside the draft strategy. It records the source behind each proof point and distinguishes evidence from inference and assumption. That makes the gaps visible: what can we substantiate, and what still needs checking? I also run parallel analyses through Claude, Codex and Gemini to compare their conclusions and catch flaws.

Writing for AI Models

This work has also changed how I develop public content and advise clients on being found and accurately represented in AI answers.

Businesses often explain themselves through their capabilities. A consultancy lists its disciplines. A technology company describes its products. The prospective customer then has to connect those descriptions to the problem they face.

When we rebuilt our website, we began with situations customers would recognise and explained how our expertise could help. Case studies and other evidence were used to support that explanation.

This is central now to how I approach answer engine optimisation, or AEO, and generative engine optimisation, or GEO. I work on making content usable as an answer and strengthening the evidence supporting its inclusion and citation.

Someone researching suppliers through an AI assistant may describe a problem without knowing which service they need. Our content should explain the relationship between that problem, our work and the results we can substantiate.

Models can draw connections. I prefer to make the important ones explicit, then check whether the assistants represent them accurately.

Through Squarespace’s AI Visibility tool, I track twelve questions across ChatGPT and Gemini. Seven name Pinpoint; five ask about choosing an agency or solving a communications problem without naming us. The checks show whether we appear and give me answers to inspect for omissions or inaccuracies.

Since relaunching the website in late May, we have received three to five highly qualified enquiries a week. Although I cannot attribute that increase to AI visibility alone, the timing, and the fact that our pipeline of leads was dry before the relaunch lends itself to that perspective. 

The benefits AI brings

I am now more productive than I have ever been in my career. AI allows me to examine more material, test more approaches and develop work in greater detail.

Yet wording needs particular attention. In a fee proposal drafted by an AI, the words “spokesperson interview preparation” could sound like an offer of formal media training. “Crisis preparedness” could imply that managing a live incident is included in a retainer when it should be charged separately. 

In long exchanges, I have also seen tools lose context and recommend changes that contradict earlier decisions. I still need to shape the thinking, edit the content and decide when further revisions have stopped helping.

The clearest benefit has been the ability to apply our methods more consistently, examine the evidence more deeply and make decisions earlier—including the decision not to pursue a piece of business.