You’ve created an amazing piece of content.
You know it’s newsworthy, and you’ve generated a list of prospects to pitch with one of your favorite tools.
You’re really worried though, because you know your content might not fit “certain world views.”
For instance, say your project is an analysis of LGBTQ+ laws in the United States, and it takes a very pro-LGBTQ+ stance.
You probably want to avoid pitching your story to sites like thedailycaller.com or drudge.com.
The worst-case scenario for pitching the wrong publisher isn’t just that your pitch will go unanswered.
In addition the risk of wasting a pitch, you can also unnecessarily damage rapport with a contact for a future potential project (that isn’t politically-related).
Time and time again, journalists report that their biggest pet peeve is when someone pitches them content that’s off their beat.
You want to be familiar with their work and therefore aware of their political leaning before you send them content.
This is where it can get tricky, especially when pitching websites and publishers that might be somewhere in the middle.
How can you know their stance on every topic?
Skilled digital PRs will have an innate sense of the leanings of publishers because of their time and experience in the field, but nobody can really know the political leaning of every author or the general political leaning of any news publication generally.
Ideally we’d have a way of automatically assessing the biases of specific articles and the aggregated bias scores for domains and authors.
It just so happens that an artificial intelligence researcher named Zach Estela created an AI model trained on a dataset of 100,000 human-labeled news articles.
Using the model, any article can be given scores for:
- Political bias (from a range of extremely left to extremely right).
- Factual reporting level (from a range of low to very high).
- Fake News.
Scores can then be aggregated and averaged for both individual writers as well as publications as a whole, giving insight into either the individual being pitched or the publication in general.
An Example Project: LGBTQ+ Discrimination in the Workplace
The above campaign asset is from a campaign we produced that explored LGBTQ+ discrimination in the workplace across the United States.
This is a good example of a situation in which outreach was limited to publishers with a particular political leaning.
Throughout the content creation process, our PR team works with our creatives executing the project to consult on the best angles for pitching, attempting to enhance:
- The content’s number of pitchable takeaways.
- The breadth of possible publisher types we could target.
Through this process, they strategize how they’ll pitch and compile a list of possible pitch targets.
In this case, we had more than 100 writers on our list from publications like Yahoo, BuzzFeed, Huffington Post, ABC News, and more.
Then, using a scraper we built, we pulled the raw text of all the articles we could find written by all of the above writers.
Each article was then fed into the bias model, and the scores were aggregated as medians for each bias category.
Below is a tableau visualization (click to explore) that shows these median values for each category for each author as well as the number of articles measured.
Having at least three articles was the threshold we set for inclusion.
Key Learnings from the Analysis
By aggregating the data output from the bias model for each author, we found some pretty interesting insights:
- Center-Left, Left, and Extreme-Left were significantly overrepresented. For this pro-LGBTQ+ article, this shows us our PR person did an overall great job of targeting journalists with a likely existing positive view of LGBTQ+ rights.
- A few authors were found to have significantly higher right-leaning, conspiracy, fake news, and hate scores.
- Sam Easter: High conservative scores.
- Mike Lacey: High hate and high conservative scores.
- Brad Polumbo: High hate and fake news scores.
- Ellen McGirt: High conservative scores.
- Dominic Holden: High hate and high conservative scores.
- Suzannah Weiss: High conspiracy scores.
- A few authors were standouts in a positive way, having low fake news, conspiracy, and hate scores:
- Julie Compton: Low on negative metric scores,
- James Cain: Low fake news, conspiracy, and hate scores. (Interesting example because some of his articles are tagged very left-leaning, while others are right-leaning. A mixed opinion centrist.)
- Additional analysis of the articles aggregated by domain could also be helpful, but with this sample, there were not enough articles for a statistically significant result. In a future article, I will show how this can be done via larger scale scraping when first beginning pitching a new niche. By running the bias detector on tens or hundreds of thousands of article text, accurate domain-level biases can be detected.
The takeaways above resulted in a stronger focus on candidate journalists whose political leanings better match the story angles (pro-LGBTQ+ – Left Leaning), while removing or devaluing the focus on journalists who are less likely to either agree with the LGBTQ+-related story or find it compelling.
While pitching a campaign, there are times when you will have time to individually review each writer’s past work and social media footprint in order to determine their political leaning.
But, especially in today’s political climate, the news cycle is ongoing and up-to-the-minute.
You can use AI to reduce the burden of doing the legwork yourself so you can pitch timely campaigns sooner, while they’re still relevant.
A data-driven approach to both bulk outreach pitch list building as well as secondary analysis on that bulk list can help PR professionals pitching content:
- Better optimize their time.
- Increase conversion rate (getting a pitched story picked up).
- Decrease the likelihood of an angry journalist getting a pitch on something they would likely never want to write about.
In the long term, this type of analysis can pay big dividends by leading to a significantly improved rolodex of publisher contacts while minimizing risk.