According to 2023 statistics, more than 10% of Americans completely stopped watching certain news outlets after they had concerns about fake news.
That’s the best introduction we could think of for a blog post on how to fact check AI content.
Seriously, though, if you run a content publishing business, ‘news sources to avoid’ is a naughty list you don’t want to be on. Not even Santa Claus himself can get you off the list.
Of course, all content must be fact-checked before publishing, but verifying information and evaluating sources assume even greater importance with AI content.
In this blog post, you will learn how to fact-check AI content so you can avoid the reputation-killing label of fake news peddler. After reading, you will be more confident in publishing quality AIO content.
To start, let’s discuss what fact checking means.
What is fact-checking in proofreading?
Fact-checking is the process of verifying facts and figures and evaluating sources to ensure they are accurate and fairly represented before you publish a piece of written content.
The fact-checking process involves researching, evaluating, and cross-referencing information from multiple sources to determine its relevance and accuracy. You can do it manually or with the help of automated tools.
Your goal as an AIO editor when you fact-check AI-generated articles is to prevent instances where readers doubt the authenticity of the information you share so much that they feel the need to fact-check it.
When done right, fact-checking is painstaking, tedious work. But the results are well worth the grunt work. In return you get exceptionally high-quality, trustable content and editorial integrity, which add up to a solid brand image.
Why is it Important to Fact-check AI content?
The primary benefit of fact-checking is that it ensures the publication of accurate information, which builds trust between readers and writers. It helps to prevent misinformation, which can cause serious harm if left unchecked.
Fact-checking allows you to back up with evidence claims that AI writing tools make. This strengthens your credibility as an authority on the subject matter you are writing about. Accurate facts and citations also enable genuine, authentic communication to take place.
The whole fact-checking process can be reduced to finding and verifying anything that a reader might suspect to be incorrect and adding the right context to the facts that the AI collects. You fact-check for the reader because you care about the quality of information you serve them.
What are the main weaknesses of AI content?
AI content has been derided for sounding robotic and unnatural, but that does not cause anyone any harm. Publishing inaccurate facts and citing dubious sources, however, can cause damage to people’s health and finances.
It does not help the credibility of AI-generated content that the one flaw that seemed to check the frenzy around ChatGPT as this human-like AI tool that was smarter than most humans was its tendency to make up facts. And do so with disarming confidence, too.
Now, even AI writing tools like Content At Scale which is trained to research and use facts from trustable sources have to convince users they are not auto-generators of fiction.
More content produced means more content to fact-check
It’s now easier than ever before to publish a website. TechJury puts the number of active websites in 2023 at 200 million. Even if anyone tried, that’s a big number to police.
The proliferation of websites has naturally led to a rise in fake news and misinformation online, any of which an AI writer can pick up and reference. AI writing tools have also made it easier to scale content production, which means content authenticity may be sacrificed for content velocity.
As you would expect, readers are now more careful about their information sources. People don’t believe everything they read these days.
It’s not just consumers who nurture a distaste for misinformation. April 2nd is now International Fact-checking Day, designated as such to spotlight the growing scourge of fake news.
The fact that it is observed a day after April Fools Day – a day even national publications can lead with a hoax or two – itself is a stark reminder that all misinformation must be quickly snuffed out!
The main reason for fact-checking AI content, though, is some AI applications can’t accurately detect factual errors due to a lack of understanding of complex language nuances found in human communication.
That being said, with the right process and resources, we can verify facts quickly and effectively. Let’s explore how to do this in more detail.
How to Fact-check AI Content, Step-by-Step
Here are the fact-checking steps to follow to make certain that your AI-produced content pieces are accurate, balanced, and up-to-date:
1. Remove fluff
This is the first step in our CRAFT framework for perfecting AI articles. Basically, here you are editing the article to remove redundant language and irrelevant information that only serve to bump up the word count.
If you have not completed this step, you should probably do that first. Fact-checking a leaner draft makes the task look more manageable and, in fact, reduces the amount of work you have to do. You do not want to devote time to verifying information you will eventually edit out.
If you had written the article yourself and are self-editing, you will probably want to take a break and step away from your computer before starting the AIO editing task. Taking a break can help to detach you from the draft so you can approach it as a fact-checker.
With this article, though, you don’t have to break because you didn’t write it. The prompt is the only thing you wrote.
I have a favorite font that I prefer to write and edit in. It puts me ‘in the zone’. If that sounds a lot like your own process, changing the font type and size in the Content At Scale app is easy. Here’s where you will find them:
Make sure the cursor is positioned in the editor and then click to make sure you are highlighting the text and not the whole webpage. CTR+A to highlight the whole text and then toggle the icons to select your preferred font type and size.
If expanding the page helps to focus your mind on the text, you can do that too as indicated in the same image above. After you have expanded the editor, it will look like this:
Now that the page looks exactly like you want it, it’s time to put on your fact-checker hat:
2. Highlight the items to verify
With your draft lean and trim, you can now put it under proper examination and highlight any facts and claims that can be disputed.
To do this, you will have to read the draft again. This time do it SLOWLY, word by word, and pausing at every full stop. For full concentration, it can help to read aloud.
As you read, you want to highlight the following:
- Names of people (and their titles), places, businesses, websites, and organizations
- Figures, including statistics, ages of people and organizations, and dates
- Sources – both those you have linked to and those you haven’t,
- Every reference to time, date, distance, location, and descriptions of physical objects and locations,
- Expletives like only, top, first, and most.
There are many ways to highlight the text you want to verify. You can bold, underline, italicize, or color the text. But the best way is to use a marker.
The CAS app has a yellow marker that you can use to highlight text. You can find it under styles on the editor settings:
Highlight the text you want to mark. Then click the styles caret to reveal a drop-down menu. Select ‘Marker’ right at the bottom. Your text should now be highlighted.
Sidenote: Did you wonder why I used the yellow and purple colors above?
3. Verify all facts
This is where the rubber meets the road. Here’s what you do to double-check and verify the facts in the article you are editing:
I should probably have said ‘all proper nouns’, which refers to names of things – people, places, buildings, cities, countries, etc.
Here you are double-checking to see if you are referencing the right person or place and if the name is spelled correctly. Once you refer to the wrong name or misspell it, the information becomes inaccurate.
Remember that some names are common. Many buildings and people share names. So instead of saying:
‘According to John Smith, a leading content marketer..’,
… it’s more helpful to the reader to say:
‘According to John Smith, who leads content marketing at XYZ Corp’
It won’t be a shock to find a few people named John Smith who are all accomplished content marketers.
For buildings and place names, adding a neighborhood, town, state, or country will make the information clearer. Do you mean:
- Moscow in Russia or Moscow in Kansas?
- Naples in Italy or Naples in Florida?
- Cambridge in England or Cambridge in Massachusetts?
- Guatemala the city or Guatemala the country?
Always assume the reader doesn’t know the person or place mentioned and be as clear as possible.
Numbers and figures
Misspelling names and places can misdirect people, but quoting the wrong figures poses far greater dangers. For example, it can exaggerate or understate prices, which can cause people to make wrong purchase decisions.
Quoting wrong percentages, speeds, sizes, and other measurements can also lead people to make the wrong conclusions on important matters. On the hand, wrong distance figures can make a destination seem farther or nearer than it really is.
Dates and time
Quoting wrong dates can make outdated information appear fresh and current, again to the detriment of readers. Incorrect times distort the information you are sharing and cause readers to doubt its authenticity.
The use of expletives in writing can be misleading. In short, if you don’t have the data to prove that something is the ‘best’, ‘cheapest’, ‘only’, ‘first of its kind’, or ‘most trusted’, remove the expletive.
Unless you can point readers to your source and the source itself is credible, don’t claim that your app is ‘the most downloaded’ in your niche. Simply state how many times it has been downloaded. If you can add a screenshot to support your figures, the better:
Take extra caution to double-check every reference to time, date, distance, location, and description of places and physical locations. Thanks to the tools that we will share below, it’s not that hard to do.
It will reflect badly on you if the reader has to fact-check you and disprove the claims your article is making, so make the effort.
4. Gut-test your conclusions
After fact-checking the entire text, give it a once-over for anything you didn’t verify. As you do that, pause where you find the article makes generalized conclusions.
Chop off any conclusions and opinions that sound biased towards a position, are outlandish, or trigger suspicions of conflict of interest. Or find the data to back it up.
We have verified the facts, how about the sources of those facts?
How to Evaluate Sources Cited in AI Content
When fact-checking any written content, it’s important to take the time to assess the credibility of the sources used. Look at:
- who wrote or published the material;
- what credentials the author has,
- when was it written;
- why the source is sharing that information;
- how current the information is.
All of these factors help determine if a source is credible. Double-check any statistics or quotes the source shares by looking at other reputable sources that provide similar information.
Let’s now discuss tips for checking if a source is credible and reliable:
7 tips for evaluating sources
Just because another website has made a claim or used a statistic in their content does not mean that information is authentic and good to quote or reference in your article.
Check to see if they mention or link to the source of their statistic or claim. If they don’t, take that information with a grain of salt. Do your own digging to find the original source if you really must use the information.
Here are more tips for evaluating sources before you use their information in your articles:
- Read the source material before using it
It is dangerous to use facts and figures from a piece of content you have not read in full. Most headlines these days are crafted to attract clicks and not to educate readers.
Read the article itself to see if the bold claim made in the headline is supported with solid facts. Clickbaity headlines rarely deliver on their promises. In fact, if the headline is clickbaity, find another source.
- Use highly cited sources
Google now has a label for Highly Cited sources on mobile search that it hopes will help readers find helpful and relevant information. As a publisher, these are good sources to cite in your content.
For breaking or major news stories, the media outlets that get the label typically have local journalists who write their stories. Their facts and figures are gathered from interviews with officials from national agencies and are, therefore, usually very reliable.
- Remove your main source from the search results
Before using any facts from a source, check to see if any other publications have cited the source. If no one else has referenced the information, dig around to verify it before you use it. Or just delete it.
An easy way to check if any other reputation publication other than your source has used the fact is to remove your source from search results for the query. You can do that with the -site:(your source’s URL) search operator.
Earlier we disproved the myth that sugar causes hyperactivity in children. When we search Google we will see a featured snippet affirming the belief is indeed myth:
Assuming that’s the source we used to debunk the myth, let’s remove it from the search results to see if there are other publications that have fact-checked the belief.
We will search for sugar causes hyperactivity in children -site:eatright.org. Here’s what Google returns:
- Separate myths from facts
Just because a fact or claim has been quoted many times does not mean it is true. This is especially true of common beliefs, many of which are just myths that have since been disproved:
- A goldfish’s memory span is much longer than the 3 seconds it’s said to be. It can be as long as 5 months.
- You are unlikely to die if hit on the head by a penny dropped from the top of the Empire State Building,
- According to Einstein himself, he did not fail math class,
- Drinking alcohol does not raise your body temperature. It lowers it.
- Sugar does not cause hyperactivity in children (see images just above).
The above ‘facts’ are myths that have been debunked by studies conducted by reputable scientific bodies. Don’t propagate them by using them in your content.
- Counterbalance facts with the source’s biases
As humans, we have inherent biases that can get in the way of facts. Because of these biases we often present information in ways that suit some goal or agenda. Scientists call it motivated cognition.
Before you use information from a source, check for any potential biases. Review its other content to identify causes and beliefs they typically lean to. Check their About Us page and social profiles to pick any themes in the content they produce and share and the voices they promote.
A common example is of national newspapers that mask downright political biases as editorial policy. Be careful using the facts and figures they share on political topics as they typically overlook relevant counter-facts.
- Check the publishing date
For topics in the ever-evolving tech niche, for example, the information a webpage contains is usually outdated and not very helpful if it’s over a year old.
A bug fix from an article suggested may no longer work if the software itself has been updated many times since the article’s publication date. So always check the publication date to save yourself from the embarrassment of being fact-checked by readers in the comments.
- Check if the images are professional
Credible sources put as much care into the text as the media in their content. They invest in original images instead of using royalty-free stock images pulled from the internet.
If a publication used a photo they took themselves, the information they are sharing about the event is probably reliable. It at least shows that they were there in person and aren’t recycling content from other sources.
We will share tools for verifying the originality of images further later.
- Verify with more than one source
One source may be wrong, but more than two are rarely wrong. Unless it’s a popular myth. If more than a couple of sources have referenced the same fact, it may be good to use.
The other reason to check other sources is that different conclusions can be reached from the same facts. Instead of quoting someone’s conclusions, go down to the data itself and make your own conclusions.
Every fact needs context for useful conclusions to be drawn. So ignore the way the source has framed their argument and ask yourself what the fact means in the context of your own piece. Dig deeper and check the sources your source is quoting from.
Appearances may be deceiving, so how else can you tell if all these sources are trustable?
Tools for Verifying Facts and Evaluating Sources
On the surface, a source can seem credible. And you may not have the time to manually trawl the internet and make phone calls and send emails to make certain you are using credible sources.
Here are online tools you can use to fact-check your AI content:
1. Google Fact Check Explorer
Google Fact Check Explorer allows you to fact-check by topic. This search engine checks all major fact-checking websites to verify facts and surface fact-checked fake news on specific topics.
Here’s what it returns for ‘2020 US Election’, a topic surrounded by much fake news and conspiracy theories:
Google Search also now has a Fact Check by feature on both mobile and desktop. If a fact has been frequently contested, there is a chance an independent fact-checking organization has investigated its authenticity:
If indeed it has been verified, the article carrying the fact-checked information will be marked with the fact-check label under its SERP listing. The publisher will, however, have to be using Schema.org ClaimReview markup.
2. About this result – Google
Sometimes checking a source’s website will not show anything fishy, but you may still want to verify with a trustable third-party site. Queue Google’s About this result feature on mobile.
Toggling the three dots icon beside the page’s SERP listing for the search query will reveal the About this result option. Clicking it prompts the algorithm to fetch the source website’s Wikipedia page.
The pop-up that appears will show a brief description of the publisher along with a link to the Wikipedia page that the snippet is pulled from so you can learn more.
3. Google Reverse Image Search
Less credible sources tend to use original images out of context or doctored ones to support their claims. You can use Google Images’ reverse image search function to check if any other outlets have used the image. If it’s been widely used, it is probably original.
UK-based Daily Mail carried this story of a lion on the loose in the streets of Johannesburg, South Africa. To convince readers the unusual incident actually happened, they would have to show a picture or video of the lion walking the streets.
If they do include a picture, you want to be sure that it is not photoshopped. After doing a Google reverse image search, I can see that the story has indeed been covered by other news organizations that have also used the same image.
4. Plagiarism checkers
Another good way to verify the credibility of a source is to check if its content is original. If there are traces of plagiarism, you have to assume their information may not be well-researched.
There are many plagiarism checkers on the market, including Grammarly, but Copyscape is the gold standard.
Copy the text from your source article and paste it into Copyscape or your preferred plagiarism checker. You want the score to show zero percent plagiarism.
Remember Copyscape is backed into the Content At Scale app, so you can easily check if the articles you produce with the tool are original. The article you are reading showed plagiarism:
Claimbuster uses machine learning, natural language processing, and a database query to separate the truth from the lies. The tool, for which you need to request a free API, delivers near-instantaneous results and can process both verbal and written fact-check requests.
Claimbuster uses search engines, and a repository of claims that have already been fact-checked by other organizations to monitor, spot, and match claims made on various kinds of media, including on social media, and even on broadcast media.
For actual fact-checking, Claimbuster compares statements against established factual evidence from authoritative sources such as Wikipedia entries or news articles from reliable outlets like The New York Times and BBC News Online.
Other fact-checking tools you can use
Popular software solutions used for automated fact-checking include Google Scholar Search Engine Optimization Toolkit (SEO) and Microsoft Word’s Natural Language Processing (NLP).
Examples of AI applications used specifically for fact-checking include:
- IBM Watson Discovery Service, which employs natural language processing technology to search through extensive databases of unstructured data;
- OpenCog Prime, which leverages machine learning algorithms to detect patterns within text documents
These tools use artificial intelligence (AI) algorithms to quickly analyze large amounts of data from multiple sources in order to detect inconsistencies or inaccuracies in a given text document or web page.
The internet provides an abundance of other resources for verifying facts and claims in AI content. Sites such as Snopes and PolitiFact offer comprehensive fact-checking services for validating information from a wide range of sources.
How to Fact-Check – Your FAQs, Answered
How can I verify the accuracy of a source?
To verify how accurate a fact is, first check for any bias or conflicts of interest that may be present in the source material. Look for evidence to support any claims made by the source and consider other sources that could provide additional information on the same topic.
What are some reliable fact-checking websites?
There are a number of reliable fact-checking websites AIO writers and editors can use, such as Snopes, PolitiFact, and FactCheck.org. These sites provide independent reviews and analyses of news stories to help ensure that readers have access to accurate information. They also offer ratings on the reliability of sources used in articles or other media content.
How do I know if a news article is credible?
Look for reputable sources that have been cited in the article. Check if the author has any credentials or expertise on the topic being discussed. Also, look at how recent the information is and whether there are any other sources confirming what was written in the article.
Be sure to read through all of the content carefully and consider if it makes sense logically. If something seems off or too good to be true, it probably isn’t credible.
Is it possible to detect bias in fact-checking reports?
You can detect bias in fact-checking reports by carefully examining the language used and the sources cited. If there are discrepancies between facts presented and sources quoted, or certain words appear to have been chosen deliberately to sway opinion, then the report may be biased.
Create Evergreen, Reader-First Long-form AIO Content with Content At Scale
Fact-checking is an essential part of the AIO editing process. It ensures that the information in your AI-generated articles is accurate and reliable, which helps maintain trust in a publication’s content.
It is an essential step in our CRAFT framework for perfecting AI content. But the framework needs a high-quality baseline (AI-generated article) to produce AIO content that is helpful to readers and which you can stick your brand behind.
There’s no better AI writer on the market today that produces nearly-ready-to-publish long-form articles than Content At Scale.
With a plugin that publishes straight to your WordPress site without leaving the app, among other features, at a barely believable price per long-form article, you get a lot more for a lot less of what you currently spend on content production.
Scale your content production faster for less with Content At Scale. Get started here.