I know that quite a lot of my social media attention this year has been on YouTube (where the video on Rachel Reeves, published yesterday, got more than 25,000 views in its first twenty-four hours, which is more than the reads here in the whole day), but this blog is not being ignored.
I had an idea that AI must now make it easy for summaries of blog posts to be generated by ChatGPT (or something like it) to then be inserted here to assist those with little time, and to persuade those who were not sure whether they wished to commit to reading an article that it might be worth their while to do so. The summaries could also, I imagined, be posted as 'mini-posts' on Facebook, Twitter, LinkedIn and elsewhere.
Andy Moyle and I got on the task, and the technology sort of exists, but we have not found anything reliable that:
- Produces accurate summaries
- Allows those summaries to be edited
- Is reasonably priced
Before spending more time on this, let me be clear that my aim is to:
- Make clear the summaries are distinct from the rest of the text. A box would be good for this purpose.
- Be useful and not misleading.
- Apply the summary when. required: some posts do not require one.
- Boost engagement.
I have done a couple of summaries this morning in the two blogs I have published so far today, here and here. The first was almost wholly ChatGPT generated in 50 words. The second required 70 words to be useful and had to be edited.
If anyone knows of something of the sort required, please let me know. I stress that I am looking for something that, ideally, integrates straight into WordPress to save time and effort and also to make formatting easier.
So, my questions are:
Would article summaries be useful at the head of a post?
- Yes (37%, 97 Votes)
- Sometimes (35%, 91 Votes)
- No (17%, 44 Votes)
- I am not sure (11%, 29 Votes)
Total Voters: 261

And:
How long should a blog summary be?
- Whatever is useful (76%, 196 Votes)
- I don't care (12%, 32 Votes)
- 50 words? (7%, 19 Votes)
- 40 words? (3%, 7 Votes)
- 70 words? (2%, 5 Votes)
Total Voters: 259

And for the record, ChatGPT summarised this post as:
You're developing a system for generating blog post summaries to boost engagement, allowing quick previews for readers. You want the summaries to be concise, accurate, and distinct from the main text, and you're looking for reliable, affordable tools that offer editable summaries. You're also seeking feedback on summary length preferences.
It also offered this analysis of the above questions:
It sounds like you're working on a solution for summarizing blog posts to increase engagement, particularly for those who may not have time to read an entire post. Summaries can be a great way to give readers a quick snapshot, drawing them into the full post if they're interested.
Regarding your questions:
-
- Would article summaries be useful at the head of a post?
- Yes: Summaries can be incredibly helpful for readers who are skimming. They provide a clear, concise preview of the content and help them decide whether it's worth reading the entire post.
- Sometimes: In cases where a blog post is short or highly focused, a summary might be redundant. For longer or more complex posts, however, they can be very useful.
- No: Rarely, some readers prefer diving right into the content without any prelude, though these users tend to be in the minority.
- I am not sure: It may be worth testing different approaches to see how your audience responds before fully committing.
- How long should a blog summary be?
- 40 words?: This could work for very short, direct posts. It gives a quick glimpse without being overwhelming, but may miss key details in more complex pieces.
- 50 words?: A good middle ground. It provides enough context to give a clearer picture while remaining succinct.
- 70 words?: Ideal for more detailed summaries. It allows for a bit more information to engage readers but might feel long for certain posts.
- Would article summaries be useful at the head of a post?
A hybrid approach might be ideal: use shorter summaries (around 40-50 words) for lighter posts and expand to 70 words for more detailed ones. You could also A/B test different lengths to see which drives the most engagement.
Thanks for reading this post.
You can share this post on social media of your choice by clicking these icons:
There are links to this blog's glossary in the above post that explain technical terms used in it. Follow them for more explanations.
You can subscribe to this blog's daily email here.
And if you would like to support this blog you can, here:
I tend to read the Blog post, so will not seek to comment on this matter.
Except to say, on educational attainment, the headline specifies UK; but the article only discusses England, as far as I can see. I do not doubt the principle applies to all, but that is not my point. The general tendency to ignore or erode Scotland’s differences is standard procedure in the Union; but here, it is a surprise!
If I am a tad grumpy I noted Angela Rayner has dismissed out-of-hand the Scottish Government protest that removing the Winter Fuel Allowance gives Scotland’s government a budgetary problem. Removing a British Government funded benefit obviously affects Scotland. Rayner’s stupid remark is the kind of imbecility we suffered from the Conservatives; making big Austerity cuts to reserved funded areas; which the Scottish Government then used what resources it had to mitigate (bedroom tax, free school meals for primary level, mitigation of benefit cap, free prescriptions, etc., etc).
The SNP claim mitigation costs £1Bn per year. The Conservatives, and now Labour then use the cost impact of mitigation to claim everything else is underfunded, the Scottish budget is lavish and the problems are SNP incompetence. This is British politics today. Utter stupidity, and crass, cheap lies that reduce the real issues to pulp.
My bad, as my sons would say
Anyone who cares about climate change should not be using AI to generate content.
By your logic I should also not use a search engine
Shall we get real here?
Sometimes, Richard, your replies can be brusque to the point of rudeness, especially when you are not on firm ground. Here are some articles describing the climate impacts of AI…
https://time.com/6987773/ai-data-centers-energy-usage-climate-change/
https://www.scientificamerican.com/article/the-ai-boom-could-use-a-shocking-amount-of-electricity/
https://www.forbes.com/sites/bethkindig/2024/06/20/ai-power-consumption-rapidly-becoming-mission-critical/
And note that there is a variety of opinion on this
Why be brusque on this?
Because if that was my concern I would not run a blog, post videos, buy equipment to do either, and so on
I can be accused of being a hypocrite on all these issues and no doubt someone will suggest I am
My point is, where’s the line? We all have to decide, but my question is, would you like me to turn this blog off? I don’t think that’s brusque. I think that’s the inference and I don’t see the gain.
It’s utter nonsense to claim the use of ‘AI’ will impact climate change in any meaningful way.
In fact, there are many ways in which the deep data analysis made possible by such technology could be used to reduce emissions and harms elsewhere.
Not to mention the fact that the chip technology and design is improving year on year so the processing power used for ‘AI’ will actually continue to fall considerably because companies really, really don’t like running hot and expensive server farms. As an easy example, Intel’s forthcoming chips plan to increase the processing capability for the LLM types of jobs five-fold over the next 2 years. All within the same power envelope. Companies developing ‘AI’-specific chips offer much greater performance in a much smaller power envelope as well.
Now, if you want to complain about the ludicrous levels of power which have been used to mine Bitcoin and other crypto currencies, I’m with you entirely, but AI? Nope.
‘AI’ is neither the solution to, or the cause of, all of life’s problems, regardless of what many Cassandras like to claim.
Thanks
The energy usage comes from the training of the LLM. The current releases of ChatGPT et al do not require much energy at all because they’re already trained (the “P” stands for “pre-trained”).
(Tongue-in-cheek comment) Is AI capable of recognising policy conflicts because if it can Scammer & Co badly need to use it!
https://www.theguardian.com/business/article/2024/aug/16/can-labour-fix-britain-modern-supply-side-economics
Morning Richard
I’m a bit conflicted here.
What problem are you trying to solve? You say that you want to assist those with little time – well if your posts can be accurately encapsulated in a 50 word nutshell, why are you writing longer ones?
Alternatively. What is the benefit of going after readers with little time – who may come to completely the wrong conclusions through skimming ( the result of which you already spend an inordinate amount of time and emotional energy dealing with trolls and blockheads). The USP of this blog is that thoughtful people engage with it, enhance or refute the arguments set forward, or read the points made fully and learn something as a result.
If someone is so time short they can only read 50 words, then they are not your audience anyway.
Wrt those who might come back later – usually each blog title and the first line is the hook that helps the reader to prioritise the importance of coming back to it – and is also useful in finding the article again when skimming as one ‘comes back later’.
I’m apparently too young for Xwitter so no longer have an account, and don’t use FB, so I’m not sure of the efficacy on those platforms.
I think your short videos and accompanying transcripts are often the smallest amount of words you can use to get your points over, so not sure yet another precis adds value.
Trouble is : I enjoy the deep engagement that you have achieved on this platform – which is only due to your efforts and the knowledge and respect for the readership shown by your contributors. I remain to be convinced that the cost benefit for the additional traffic from people who are time poor will be worth your time or our eyeballs. 😉
Thanks for your comment, which I appreciate.
I think that you are probably coming to this issue as someone who is relatively time rich. Many people are not. There is a widely acknowledged problem of people wishing to understand what a paper or article might be saying but not having the time to explore it in sufficient depth. This is why, for example, abstracts are attached to academic papers. They do not undermine the paper itself, and reference it, but what they do is provide insight as to whether to voting time to reading it will be a worthwhile exercise for the reader.
On Twitter, at least, this is called a TLDR, which stands for ‘too long, did not read’, meaning that a summary is required.
There are approximately 20,000 readers here a day now. Not all of them are time rich. Some leaves quite quickly. My aim is to ensure that they get at least something value from their visit.
And as for the trolls, they deliberately misunderstand and there is nothing that can be done about that. However, having a summary might show that they have done so.
But thank you for requiring that I explain that.
The problem with academic abstracts (and even titles; the standard now is a two-tier title – so some may never go past the title, because the title is an abstracted abstract). All this just encourages the curse of ‘silo’ thinking in academia, in my opinion. I have tried to insist on my own deliberately more enigmatic titles, and have not always won.
They don’t always work – as experience has proved to me
I still don’t think I am a good title writer
“I still don’t think I am a good title writer”
But you are excellent at short subject video which is the “Nuts & Bolts” of the “game” you are engaging in.
Thank you
“I enjoy the deep engagement that you have achieved on this platform”
Me too!!!!
Thanks, again
I tend to use the post title as the summary. Your introductory paragraph then provides clarification. In which case, you don’t need an additional summary.
Interesting
But the polling suggests otherwise
Social Media surfers need a summary.
I am assuming you goal is to drive more people to your sites and gain more loyal followers.
Yes
Of course
That drives the message
And then if they ate any good they become posts on Facebook, LinkedIn, etc. as well – and that makes them worthwhile
If it helps widen the impact of your daily offerings, and is relatively easy and quick (and inexpensive) to generate, then why not? Horses for courses, let a thousand flowers bloom!
Personally I found the two GPT generated summaries rather clunky and difficult to read, and I’m still trying to analyse why that might be. The sentences are generally long, stringing together several thoughts or conclusions, which is clearly the reason for summarising in the first place. I found I had to read each sentence twice, (my bad!) and that the text lacked rhythm and balance.
In short, it comes across as sterile and, dare I say it, robotic.
Reading the equivalent in your full version, I felt my interested piqued, and could engage with the subject matter much more easily.
That interesting….
I would certainly only do this if I can edit the summaries
I would not use them on all blogs
As a speed reader I find it much quicker to read your blogs that have to listen to it being spoken.
At first I felt the exact same way. Now I addicted to both the Vlogs and the written posts.
Written posts are great if you are in a situation, like a lunch break or clock watching at the end of the work day, where you cannot listen to the audio.
I struggle a bit to imagine how long 40 words or 70 words would be. How long is the ChatGPT summary in this blog, immediately after the vote? That looks a useful length, although it actually feels somewhat patronising.
The short one is 50 words
The other 200+
In principle, I would welcome summaries on some of your longer or more technical posts. However, I have just read the two on your earlier posts and I found them, short as they were, dry and rather off-putting. They have no trace of your personality, which is what gives life to your posts for me. I suspect someone who’s just flitting by would not be tempted to linger after reading them. I don’t know if it’s possible to get the AI to adopt a more engaging tone? I know ChatGPT will write poetry on any subject, so perhaps you could ask it to provide a summary in haiku form!
I know there is some summarising software that lets a style be chosen
But it will always be machine generated
I prefer to read the whole blog, but summaries can be useful for many people. In other places I read a summary, or the first paragraph, then decide whether or not to keep reading.
I’d say add them as it will likely grow your readership.
Thanks
The decision has been made
Andy is writing the code now….
Just a thought:
When you do the writing, hopefully, the written words will convey your thoughts in a meaningful way. You have control.
It would concern me that, if I were in your position, the machine generated piece would, in some way be inaccurate, and I may not notice – particularly when I was under time pressure or feeling off-colour (or worse).
Saying the machine wrote it, would be no defence….
I am not ever going to post automatically
I would always edit
Working in a media environment there is an on-going discussion around how we use AI tools. We are surprised how quickly we are adopting, but aware to the limitations, not least with specialist content.
I’d find summaries of your occasional long posts helpful. However I come to this blog because many you daily posts are already summaries of reports and articles, with some context and commentary. You get to the point upfront making it quick to filter content.
ChatGPT is useful for summarising long reports and converting the style of a written piece, or talk, for a different audience, use, context or media. Results can be very sensitive to the prompts given, and you often need to interrogate responses or as additional questions to refine the outputs. Doing that several times still seems disconcerting after a year of use, such is the speed of response.
ChatGTP can also be effective at suggesting ideas for ‘sticky’ headines, YouTube titles etc. long publishers it’s being used to analysis content and identify what’s currently working (quality of engagement) and suggest why.
Long, very specific prompts where you state the ‘who’, ‘what’, ‘where’, ‘for whom’, ‘why’ etc. seem to work best. The prompts can be given in a very conversational style thanks to the underlying language model. Indeed you can enable voice control on both ChatGPT and CoPilot and talk….rather than type.
Thanks Rod
The prototype is in development to do this as there seems to be nothing out there on the market
I will learn how to use the prompts
I can see myself spending time watching cricket and testing that….
I would imagine that people who follow this blog are self-selected as being ones who are looking for depth and informed understanding. Not a superficial summary.
However using machine-generated summaries could be useful if you want to post something on other social media (Twitter, Mastodon etc) alerting people to a new blog post.
Looking at the examples you posted today, I do see the company accounts summary gives a reasonable flavour of your ideas though as someone above says it somehow reads a bit clunkily. The A level one doesn’t really add any value for me, a single sentence (which could have been the title) would do instead: A level results again demonstrate the educational inequalities currently present in the UK, which the government should address.
You may be right Jonathan
But I am also aware that the commentators here are not representative of all the readers and so these are largely for the silent majority
The thing is—if summaries turn out to be a not-so-great idea, you can always stop generating them. You won’t know if this new approach works until you give it a go—so more power to your arm. I’m interested to see how it shapes up.
Personally, as long as the summaries accurately reflect the content of your post, I can’t see a problem.
Thanks
The point is anyone can skip them
Less is more. A concise headline (the title of the blog post) then a short paragraph introducing the post. Write it yourself after you have written the blog and don’t rely on AI.
Also maybe fewer posts each day? It’s hard to keep up with your output!
Whatever you decide I will always appreciate the time and effort you put into these posts – thank you!
The traffic figures suggest more is better here, and there are already four ideas for the morning. Not all will muse it though.
And I might use AI, but only if I like it. I bet one will go unedited
And, thanks for your comment