How the Digital PR Industry is Using AI

“Using AI can aid your planning and outreach immensely if you use it to replace lengthy desk research tasks. But we're not just talking about ChatGPT here folks. Don't sleep on the =AI() function on sheets for your media lists!! It's great for automatically summarising information about the titles or journalists in your media lists, from outlining their key topic areas to showing a journo's usual headline styles.”

“We use AI as an aid to come up with engaging subject lines, expand on creative ideas following ideation sessions (such as finding data sources) and each morning ask it to summarise news stories across a range of topics where there could be opportunities for our clients to feature.

“For example, we use the prompt 'find me the top UK automotive news stories from this morning that my car insurance client could contribute to.' We can then take this list, read each article and make the decision on how we can provide comments or data. This quick use of AI greatly speeds up our reactive work every day.”

“Outreach is no longer just about sending a well-written pitch - it’s about relevance. We use AI to analyse journalists’ historic coverage, tonal preferences, recurring themes, and even headline structures. That insight allows us to tailor intros, angles, and subject lines in a way that feels genuinely aligned with what they cover, and they are interested in. We also use AI to test and refine subject lines and adapt pitches by title vertical. This level of structured personalisation increases response rates while maintaining quality control.”

“I find AI super helpful in preparing for ideation sessions. Finding suitable reports we can use as a rationale for a campaign idea, helping identify available data sources for potential methodology, coming up with specific media angles and formats. It's a really good starting point to get ideas and conversations going before the wider session with the team where the true magic happens.”

“NotebookLM is an efficient tool for in-depth analysis of large datasets. You can feed it raw data sources such as CSVs, Whitepapers, or Survey Results, and have it identify the most surprising or headline-worthy key takeaways for Digital PR stories”

“Before AI, analysing large volumes of data for a PR campaign meant hours, sometimes days, of manual work. Sifting through reviews, categorising sentiment, and trying to spot patterns across thousands of data points was time-consuming and prone to human error. 

“For a recent campaign, I analysed over 100,000 traveller reviews, something that would have taken days of manual work and simply couldn't have been done as effectively before. A combination of AI, automation, and data engineering at scale allowed us to automatically detect sentiment, understand the context behind what people were saying (so "not worth it" was correctly read as negative, not positive), and even analyse reviews written in different languages using the Google Translate API.

“AI is particularly powerful in surfacing hidden patterns within large or messy datasets. In several campaigns, it has helped us identify unexpected correlations or regional contrasts that became the actual headline story - elevating what could have been a generic data piece into a genuinely publishable, relevant narrative. The result is stronger story construction and a higher likelihood of higher authoritative links and coverage.”

“I especially find using AI really helpful when I'm analysing survey results or large datasets. Often looking at survey data can feel intimidating, especially when there's data covering many age groups and regions, using an AI tool can save hours at analysing the results. The same applies with large datasets for campaigns, using AI can quickly pull out key themes, anomalies and potential stats much faster than manual review.”

“Using AI for survey analysis has been a game-changer. Previously, I'd rely on Excel and pivot tables to manually crunch through responses, a slow and often limiting process. Now, I've built a custom script powered by AI that processes survey data in a fraction of the time, identifying patterns and themes across hundreds of responses far more accurately than manual analysis ever could.”

“To move beyond generic AI suggestions, create a custom agent designed specifically for high-level ideation and 'stress-testing' your concepts. Start by building a Knowledge Base, using a tool like NotebookLM or a custom GPT, and upload your agency’s proprietary 'Source Bank,' such as historical Campaign Master Sheets, successful outreach tips, industry newsletters and client guidelines. Once your agent is grounded in this data, you can move from simple brainstorming to strategic validation.”

"We started experimenting with asking AI to collate the information itself, but it's not up to that task and will often hallucinate results.

However, asking it to write Python scripts to scrape information from the internet has proven much more successful, even starting with zero coding knowledge. We used it recently to scrape data on 400 films that was only available on individual pages, so would have taken several days to do manually. It also happens to be data that changes fairly regularly, so we can just re-run the scraper to be sure we've got the most up to date results, rather than either rushing to use it while it's still accurate, or having to start from scratch with manual collection."

“In the ideation phase, AI allows us to pressure test angles before we take them to market. We use it to interrogate datasets, stress test headlines, identify potential media objections, and explore alternative framings of the same core insight. Instead of settling on the first ‘good’ idea, we can rapidly explore multiple creative territories - from data-led hooks to culturally reactive angles - and evaluate which has the strongest news value.”

"We’re benchmarking how visible clients are in AI search platforms and LLMs. We all know the tools available for this are still quite unreliable right now, so we’re doing a lot of manual prompt testing to see whether our clients actually surface in AI-generated responses.

"One of my favourite tricks during this process is to ask why it didn’t include our client. The answers can be really useful, and have helped us make practical recommendations, from updating Wikipedia pages to refining company positioning, to increase the chances of appearing when potential customers use LLMs such as ChatGPT in the evaluation stages of their search. 

"At this stage, we can also reverse engineer the sources being used in AI responses. By analysing the publications that are cited, we can build a list of publications and articles we want to be featured in - think Link gap 2.0!

"This research stage is essential because it helps us understand how AI models research a topic. Bernard Huang explained this really well on the Buzzstream podcast, describing three layers in how AI generates answers: 

"the foundational layer (training data), the grounding layer (where models check and validate current information), and the personalisation/context layer.

"For Digital PR, the opportunity is to intersect with the model during the grounding layer, by creating content and coverage that aligns with the sources AI systems use to verify information.”

"A great example where you can see this in action is our recent work with Emma - The Sleep Company. By running this type of analysis, the Guardian was identified as a priority publication due to its authority, reach and influence. We secured product-led editorial coverage where Emma’s weighted blanket was ranked the best in the UK. 

"The impact was immediate. Within 24 hours of publication, Emma’s weighted blanket surfaced within Google’s AI Mode recommendation, citing The Guardian article as the source. The article also ranked #1 on Google for ‘best weighted blankets’, where no AI overview was present, making the recommendation even more powerful. 

"We also saw the impact across LLM-powered platforms. When testing a range of buyer-intent prompts within ChatGPT, Emma’s weighted blanket surfaced as the top recommended option, again referencing The Guardian article as the primary source."