GPT filling in usecase database

How to use GPT-4 for SEO Content creation?

Posted on by

GPT4 dropped last week and after its amazing product announcement, we know it will change the world even more. But how useful is it for streamlining content ops? Now that using AI is no longer subject to blanket bans, since On 8th February, Google announced on its Search Central blog that it’s going to be “rewarding high quality content, however it’s produced“. Still, I’m not planning to use GPT-4 for writing entire posts – its outputs often lack nuance, organization, and depth. But there are some useful ways how I’m planning to use it.

Wanna get the exclusive programmaticSEO Templates included in this post, access to a content ops community and much more?

Join my Content Operations Patreon membership! 

content ops for saas

content ops for saas

1. Outlines for blog post briefs

Outlines for blog post briefs built with GPT

Creating briefs for SEO blog posts (example of a good brief template here) is a long, expensive and quite honestly – exhausting process. It is necessary though if you want to maintain high quality of your posts, esp BOFU posts for B2B SaaS – when you’re scaling your content operations.  Looks like GPT4 can help to speed things up a lot by creating outlines, which can be a great starting point for such brief:

Here’s an example of a blog post outline ChatGPT created for me in 1:41, based on this prompt:

Write me an outline of an SEO-optimized blog post for the topic: “The skyscraper link building  – the Ultimate Guide” knowing that: 1) The primary keyword is “skyscraper link building” 2) You need to include all the H2s, H3s and H4s based on top-ranking SERPs in the correct order 3) You need to include the key talking points for each paragraph in bullet points 4) include links to any resources that might be helpful for the writer to read more about the topic for each paragraph. 5) Include Introduction and Conclusion.

It took GPT-4 1 minute 40 seconds to create *this* outline:

Title: The Skyscraper Link Building – The Ultimate Guide

I. Introduction

II. H2: What is Skyscraper Link Building?

III. H2: Finding the Right Content to Outrank

IV. H2: Creating Outstanding Content

V. H2: Promoting Your Skyscraper Content

  • H3: Building Relationships with Influencers and AuthoritiesThe importance of building a network in your niche
    Reaching out to influencers and authorities to share your content
    Tools for finding influencers: BuzzStream, NinjaOutreach, and Followerwonk
    Resource: https://www.searchenginejournal.com/influencer-outreach-guide/384067/
  • H3: Effective Outreach Strategies for Skyscraper Link BuildingPersonalized email outreach
    Social media promotion and engagement
    Guest posting and content syndication
    Resource: https://backlinko.com/link-building-strategies
  • H3: Tracking and Measuring SuccessMonitoring backlinks and website traffic
  • H3: Conclusion

GPT 4 brief

Not bad for less than 2 minutes work!

But of course – such outline would still require some work to turn it into a fully-fledged brief:

– Cross-referencing this outline with the headings of the top-ranking posts in SERPs

– Adding product-specific content into specific paragraphs – this is what AI absolutely *can’t* do as of now. If you’re writing a BOFU piece for a B2B SaaS, you want to anchor your product in the context of the keyword you’re optimizing your blog for. Even less-experienced human editors often can’t do it well. If you miss this part – you’re missing out on the post’s conversion/monetisation potential.

– Adding CTAs and images – as of now, the paid version of ChatGPT, which allows you to use the GPT4 model – does not include visual assets in the text. And these are crucial for If you have

Here’s a brief, for example:

Userpilot's brief example

And here’s the GPT-4 outline for the same topic:

Prompt: Write me an outline of an SEO-optimized blog post for the topic: “How to Analyze NPS Responses and Get Actionable Insights That Help Your Business” knowing that: 1) The primary keyword is “how to analyze nps responses” 2) You need to include all the H2s, H3s and H4s based on top-ranking SERPs in the correct order 3) You need to include the key talking points for each paragraph in bullet points 4) include links to any resources that might be helpful for the writer to read more about the topic for each paragraph. 5) Include Introduction and Conclusion.

Title: How to Analyze NPS Responses and Get Actionable Insights That Help Your Business

Introduction:

  • Explain the importance of Net Promoter Score (NPS) for businesses
  • Introduce the main goal: learning how to analyze NPS responses and extracting valuable insights
  • Mention the primary keyword: how to analyze NPS responses

H2: Understanding Net Promoter Score (NPS)

H2: How to Analyze NPS Responses

  • H3: Organizing and segmenting NPS dataExplain the importance of organizing and segmenting data
    Provide examples of segmentation (e.g., demographics, product/service type)
  • H3: Identifying trends and patternsDiscuss the importance of spotting trends in NPS responses
    Explain how to use various visualization tools to identify patterns
    Link: Data Visualization for NPS
  • H3: Qualitative analysis of open-ended feedbackExplain the significance of analyzing open-ended feedback
    Discuss text analysis techniques like sentiment analysis and topic modeling
    Link: Qualitative NPS Analysis

H2: Turning NPS Insights into Actionable Steps

  • H3: Addressing negative feedbackExplain how to approach and resolve customer concerns based on Detractor feedback
    Discuss the importance of addressing negative feedback quickly
  • H3: Leveraging positive feedbackExplain how to use Promoter feedback to identify strengths and improve marketing efforts
    Discuss ways to encourage word-of-mouth referrals from Promoters
  • H3: Engaging with Passives to turn them into PromotersDiscuss strategies for engaging with Passives to improve their experience
    Explain the importance of continuous improvement to turn Passives into Promoters

H2: Monitoring and Adjusting Your NPS Strategy

  • H3: Tracking NPS over timeExplain the importance of monitoring NPS trends over time
    Discuss benchmarks and best practices for NPS tracking
    Link: NPS Benchmarks
  • H3: Continuously improving customer experienceDiscuss the ongoing process of refining and enhancing customer experience
    Explain how regular NPS analysis and action can lead to business growth

Conclusion:

  • Recap the importance of learning how to analyze NPS responses
  • Highlight the actionable insights gained from NPS analysis
  • Encourage readers to implement NPS analysis and optimization in their businesses

VERDICT: for generating outlines (not briefs!) of a blog post – it’s a yes (but will require some manual editing).

2. Ask GPT-4 to create a TL;DR for your posts

This is something I’m definitely going to use a lot, and already used on a couple of our programmatic posts. If you have been following my programmatic post journey (e.g. here’s an example of a Google Docs template to generate a post following the “Best {modifier} [company type] in [city/state/country]” pattern) – you know that so far, I haven’t figured out how to create the TL;DR section programmatically.

TL;DR is essentially a summary of the key takeaways from the post – so the reader can get the gist without spending 10 minutes reading the whole damn thing.

Cue: they don’t do it anyway! So if you write the TL;DR summary, and include CTAs to check out your product there – bingo! You’ve just massively increased your chance for conversion.

Not to mention this is what the users actually want (I’m guilty of reading just the TL;DRs of many posts on our blog…) – so it has a positive impact on the UX, and as we know – Google loves good UX on an SEO blog post…

Here’s an example of a TL;DR I recently wrote using GPT4:

TLDR example

This was for a programmatically generated blog post, and I knew the content – just couldn’t be bothered to sum it up in nifty bullet points. So I asked GPT4 to do it – and it did great!

Here’s how to produce good TL;DRs with ChatGPT (GPT4):

– break your post down into 3-4 chunks, depending on the length – ChatGPT won’t summarize an entire post of 4,000 words, because of the input limit. Break it  based on the variables / other organizational “aspect” – e.g. for a post on “Chameleon vs. Userguiding – which is better for interactive user guides?”

The variables in this post are:

– interactive user guides

– Chameleon

– Userguiding

– Our tool (Userpilot)

So I fed ChatGPT fragments of the post related to these variables (in this order – makes sense to introduce the main criterion of comparison before comparing the two tools, doesn’t it!) and asked GPT-4 to summarize it in bullet points using this prompt:

Summarize this part of a blog post in 5-7 bullet points (unordered list). Write it in full sentences (still in bullet points) in a more conversational and human style

Here’s the output for “Chameleon”:

chameleon TLDR

I then combined all three summaries into one summary.

Now! Don’t forget to add internal links to relevant articles and CTA to try out your tool into your TL;DR.

Tip: if the ‘combined’ TL;DR is too long, you can ask GPT4 to summarize it, say, in only 10 points:

Prompt: Summarize this text in 10-15 bullet points, in a human, conversational style, without losing meaning of any of the points below: Interactive user guides improve onboarding and feature adoption.[add the TL;DR you want to summarize]

❗ now, always do read the output – ChatGPT (even GPT-4…) sometimes come up with things that weren’t even in the prompt…

GPT 4 not following the prompt

To avoid such boopers, you can also do the same using Wordtune: this tool allows you to both summarize and paraphrase the text. Wordtune uses a different language processing model (HAIM) so it bypasses GPT-based AI detection algorithms (if you fear Google might object to too much GPT involvement):

text shortening in Wordtune

VERDICT: for generating TL;DRs from a blog post – it’s a yes (but will require some manual editing and careful attention).

3. Use GPT-4 to create comparison tables for your BoFu Tool Comparison posts

Comparison table enhance user experience on your blog and increase the chance your user is *actually* going to read the content of your post and walk away with the conclusion that your tool is the best.

Here’s an example of such table from one of our posts, made by our graphic designer:

Userpilot table

The issue is, these are pretty challenging to make and you need a graphic designer in the first place. What if you don’t have one on your team?

I tested if ChatGPT could come in handy in that case.

I used the same post as for the TL;DR – Userguiding vs Chameleon – which is better for interactive user guides? – and the previously generated TL;DR – to ask GPT-4 to create a comparison table for me.

This is the prompt I used and the output I got:

Userpilot GPT4 prompt

Prompt:

Based on the text comparing the 3 tools below, create and fill a comparison table (in HTML) entitled “Userpilot vs. Userguiding vs. Chameleon for interactive guides” with the following header rows: 1) Tool 2) General Tool Description 3) Pros 4) Cons 5) Pricing. Start from Userpilot. Here’s the text you can use to fill in the table: [I inserted the TL;DR here]

GPT-4 Output:

HTML output

The table was good, but didn’t look too pretty:

GPT4 table output

…so I asked it to style it:

table styling in chatGPT

Prompt: Style it differently: make the border lines solid, make header rows in #eb7c9b and font bold. Rearrange the order of tools to be 1) Userpilot 2) Chameleon 3) Userguiding

Here’s what the output looks like now: Ooops.

HTML table generated by GPT4

It doesn’t look that much better, but it did change the order of the tools. But to be fair, I didn’t give it the right prompt. Let’s ask it *again*:

Prompt: use this color: #eb7c9b  for header row, make the border: 3px solid, make the row lines 1 px solid

And it failed again:

VERDICT: for generating comparison tables from TL;DRs from a blog post – it’s a yes  – but you won’t be able to style it without knowledge of HTML and CSS anyway.

4. Use ChatGPT to fill in your programmatic SEO database – “Use case” database (the more ‘generic’ paragraphs for each variable)

As you remember, I’m generating some posts (“programmatic posts“) using Google Sheet templates and linked databases also in Google Sheets:

Programmatic SEO database

The “Use case” database contains a list of keywords that are used in posts like “Best tools for {use case}” or “{Tool1} vs. {Tool2} – which is better for {Use case}” – you can find examples of these templates on this Patreon as well.

The keywords tend to be quite generic, so why not use GPT-4 To generate them?

I first tried to see how it would perform for KW “no-code growth” – I wanted to generate content for “What is {no-code growth” and “Why is {no-code growth} important” database entries, in HTML (as the entries are then uploaded straight to the code editor in the blog.)

These are the prompts I used:

1) What is no-code growth? Provide a definition in one paragraph, formatted in HTML. write it in the context of B2B SaaS, and growth here means “Product-Led Growth”

2) Now write a paragraph on why is no-code growth important? Give me 1 introductory sentence and then the reasons in bullet points – format it all in HTML. Write in a friendly, informal style – this is for a blog.

It did pretty well!

I’d still touch it up a bit (manually – unfortunately it’s not possible to rewrite HTML code in Wordtune:

Now, you can generate entries with GPT-4 and paste them manually, or generate the responses directly in your database (=Google Sheet). Unfortunately, I don’t have GPT-4 developers API access (you can apply for it here), so I can only do it using “da-vinci-003” model or  “gpt-3.5-turbo”.

👉 btw. I explained in more detail “How to use ChatGPT inside your Google Sheets Progammatic SEO database?” in this post, but here’s a quick recap:

(or watch this video recap) 

– 1. Install GPT for Sheets plugin on your GoogleSheets

– 2. Enable the plugin in your sheet database + paste the API key

– 3. Go into Extensions, then GPT for Sheets > and then:

1) Enable GPT functions

2) click on “Set API key”

…and paste your API key from OpenAI account there.

– 4. 3) Now test your chat GPT on a copy of your Template Sheet (which you can make here)

❗ make sure you’ve enabled the extension on this sheet as well!

Ok, so now – what did I do? I created a Programmatic SEO database and created the prompts to generate some of the generic paragraphs for each variable (general description, why is it important, what are the types, benefits of X, why does one need tools for X, and what are the key features of these tools).

Then, in the output cells, I pasted this formula: =GPT(C3,$A4,0,100,”gpt-3.5-turbo”)

Here’s how the formula for filling in programmatic SEO databases works:

Now, I just need to drag the formula horizontally to fill in the remaining cells:

Again, it did *pretty well* – I wouldn’t use the output outright, but if you have a *large* volume of use cases you wanna cover, or you’re building an affiliate site – this can save you tons of time:

VERDICT: for general paragraphs in the “use case” database, such as “definitions”, “why it’s important”, “how to improve” etc. – it’s a yes (but will require some manual editing).

5. *DO NOT* Ask it to fill in your programmatic SEO database – a “tools” database

(But if you do have the Developers’ API access, you could probably use it based on selected existing content + a prompt listing all the columns you want to fill)

No, where it flopped: the tools database is *a lot* more difficult and time-consuming to fill than the use-case database – you’re probably not surprised by this.

Knowing how one tool is different from another typically takes some hands-on empirical experience with the tool.

use case database programmatic SEO

But what’s killing me is that I’ve already created *tons* of comparison posts based on actual experience after signing in to trials…and now finding the relevant paragraphs across 1200+ posts is a proper nightmare.

I wanted to see if ChatGPT would allow me to do that. I started from listing the columns I want to fill to save my GPT-4 query limit:

1. What is [tool] (General Description)

2. What are the pros of using it?

3. What are the cons of using it?

4. What is the pricing of the tool?

5. Positive review of the tool

6. Negative review of the tool

7. How can this tool be used for user onboarding?

8. What are these tool’s analytics features?

9.  How can I collect user feedback with this tool?

My prompt included URL sources (which unfortunately the GPT-4 model within ChatGPT can’t read yet – so I only confirmed; but again – if you have the API access, you could probably do it!)

chat GPT writing post

The output was very generic (like most content in SERPs…) and either didn’t include key differentiators, or included downright wrong information.

VERDICT: for tools database, it’s a definite no. We need to remember that GPT doesn’t think, doesn’t have experiences, and can’t read and mindfully interpret our content. It just predicts the next most likely string of tokens. This is not enough to provide in-depth analyses of key differentiators between different tools at this stage.

6. Can I use Chat GPT (GPT-4) to create a video script for me (so you can embed it in your blog)?

Now, I have this task on my weekly to-do list: I need to create a script for a few  videos based on an annual report I have. Can chatGPT do it for me? Yes it can:

Prompt: Create a video script for me based on this fragment of an industry report:

[fragment of the report]

Chat GPT (GPT-4) video script prompt

Why videos are  important for your SEO content?

Google loves interactive content. Adding videos to your blogs, especially under the TL;DR section, can potentially boost your position in SERPs. But creating edited videos from scratch is hard and time-consuming.

But what if you produce videos by asking GPT-4 to summarize and then produce a video script from the TL;DR of you blog post, and then use AI tools like Pictory or Synthesia to produce the video? 🤔 Let’s see if that would work!

Let’s create a video on “How to improve product adoption rate”

STEP 1: Take a TL;DR from your blog post on that topic, e.g. https://userpilot.com/blog/improve-product-adoption-rate-saas/

TLDR

Put it in ChatGPT with the same prompt as before:

Chat GPT (GPT-4) video script

It requires some touch ups so the AI tool doesn’t include the metatext in the video 😬

Now go to Pictory and choose “Text to video” (you can also choose “link to video” and compare results – they have a free trial so that shouldn’t be too expensive – maybe you don’t need that script at all!)

Pictory video creation

This is how it looks:

Pictory video script editor

It takes Pictory a couple of minutes to create a video:

Pictory video creation

And voila – it did a pretty good job: it even took the right visuals e.g. for the product adoption rate from some stocks:

Pictory video creation

Now let’s see how it does from the URL alone: again, Pictory takes a few minutes to ramp up:

Ok, I’d say Pictory’s script from the URL is (surprisingly, I must admit) better than the one generated by GPT-4 from the TL;DR:

Pictory video script editing

But then the final output had a ton of stock footage that just didn’t make sense:

Pictory video stock footage

VERDICT: for generating scripts for humans to record, it’s a yes. For generating whole videos from URL without any editing – it’s a not. I’d love to still explore what I can do with Pictory to edit this video.

Conclusion from this thought-experiments:

In sum, I’d happily use GPT-4 for:

– summarizing content, like for TL;DRs for posts

– creating comparison tables for BoFu blog posts comparing several variables, e.g. tools

– writing blog post outlines (but not briefs)

– filling in generic use case database entries, like e.g. metic definitions, why is the {variable} important, how to improve it, what are the benefits of…{method, tool type}, what are the key features of a {tool type} – even directly in the database, using GPT-3

– generating video scripts

What I wouldn’t use it for:

– filling in programmatic databases featuring tools

– writing whole posts

Hope you found this post useful! Let me know how you use GPT-4 for your SEO content/ content ops!

Posted in Uncategorized.

Emilia

View posts by Emilia

Emilia is a passionate SaaS marketer specializing in content marketing. She's currently the Head of Marketing at Userpilot, a Product Growth Platform for SaaS.

Related Posts

  1. Content Budget Tracker – by writer & post type

    07/09/2023
  2. SurferAI PES article

    SurferAI review – I used SurferSEO’s AI writing tool to write content & here’s the outcome

    05/14/2023
  3. GPT filling in usecase database

    Quality human content vs. AI content – what’s the difference, and how to edit the AI content to improve it?

    04/23/2023