ChatGPT was released and created by OpenAI, a Microsoft-backed company. ChatGPT has been covered extensively across the SEO community and brings up many questions about its place within the practice of SEO, content development, and Google Search.
We’ve found ChatGPT and AI technology helpful for our day-to-day SEO practices and scaling SEO strategies for clients. In this article, we explore SEO considerations, use cases, and our collective direction on using ChatGPT with SEO in mind.
Disclaimer: AI is a rapidly changing technology. Since we’ve written this article, there has been the release of Google Bard among many other popular AI tools, products, and services. ChatGPT has also been upgraded to new versions. We’ll continue to update our content periodically!
What is ChatGPT?
ChatGPT is an artificial intelligence (AI) chatbot that can take directions and provide real-time responses. It is a large language model (LLM) that can interact in conversational dialogue and provide seemingly human-curated responses.
ChatGPT currently uses GPT-3.5, which has 175 billion parameters and was trained on 570 gigabytes of text (for context, its predecessor GPT-2 has only 1.5 billion parameters). While new accounts are on a waitlist, ChatGPT is free to use during its “research preview” release.
How can ChatGPT be used for SEO?
ChatGPT can be a powerful research tool for SEO and content marketing. Currently, we’re viewing ChatGPT as a research tool for specific SEO tasks, tactics, and content inspiration. Common use cases include:
- Keyword research: ChatGPT can return related keywords, cluster keywords by topic, and translate keywords.
- Content research: while ChatGPT is great at creating content on the fly, it’s also excellent at research for content development and creation. ChatGPT can be helpful for generating FAQs, understanding content sentiment, rephrasing content, and summarizing (or condensing) existing content.
- Metadata generation: ChatGPT can provide title tags and meta descriptions for web pages based on URL or topic.
- Image optimization: we’ve used workbooks and various APIs to create scripts to automate image alt-text optimization based on the image source URL.
- Structured data: ChatGPT can return structured data markup for web pages in microformat and JSON-LD. While the code creation has been largely accurate, verifying information and validating the markup is always recommended.
- Geo-targeting tagging: ChatGPT can be used for localization with translation and geo-targeting tagging, such as hreflang markup. As with any code creation, reviewing and validating responses is always recommended.
- Reporting: ChatGPT can be helpful when creating complex formulas and regexes for reporting for Google Search Console (GSC) and custom reporting in Google Analytics.
For many of our clients, ChatGPT has been a helpful tool to scale content across many different page types and regions. We’ve successfully tested AI content creation with e-commerce websites and SaaS platforms.
Currently, the API still uses GPT-2, an outdated version, and the content returned from GPT-2 is of much lower quality, so we do not recommend using the API now. Conifr will continue monitoring API releases for the upgrade to GPT-3.5 and will explore scalable solutions.
Downsides of using ChatGPT
- It may provide outdated information: GPT-3.5 uses 2021 data from the web and may not have the most up-to-date information in its database.
- It’s not always right: GPT-3.5 collects information from the web and may not always be factually correct in its responses.
- Other users may be able to receive the same responses: technically if another user were to use the same inputs, they may receive the same responses.
- It has built-in biases: while ChatGPT may return toxic answers, the goal is to provide truthful, helpful, and harmless responses, which creates an inherent bias.
- It’s often at capacity: it’s not always reliable as an “on-demand” research tool as it’s often at capacity and unavailable for login.
What’s Google’s stance on AI-generated content?
Historically, Google has classified AI-generated content as spam and against their guidelines as automatically-generated content (as recent as November 2022). However, Google recently changed its stance to specify that AI-generated content is OK but needs to be helpful and informative to users.
Google now says it will target spammy automatically-generated content created for search engines (rather than people).
To ensure AI-generated content complies with Google’s guidelines and is less impactful during algorithmic updates, it’s recommended to follow Google’s guidelines on people-first content.
Regardless if the content is human-generated or AI-generated, Google advises brands to avoid creating content for search engines.
There are many tools publicly available to detect AI-generated content, such as Writer.com and CopyLeaks.com. If these tools can detect AI-generated content easily, then there is a high likelihood that Google can also understand the distinction between AI-generated and human-curated content.
There has also been a discussion online about OpenAI using watermarking to detect content created by ChatGPT.
For brands using AI-generated content, we strongly recommend using a human editor to ensure the content is natural sounding, relevant, high-quality, and unique to the topic. The goal is to provide unique and helpful content around any topic published on a webpage.
Testing the usage of AI-generated content may help brands understand the ranking impacts and engagement with AI content.
ChatGPT is a powerful research tool that can aid in many day-to-day SEO tasks and tactics. Understanding the limitations of ChatGPT and Google’s guidelines for helpful content is important.
If brands use ChatGPT to generate publicly available content, it’s important to use a human editor to ensure content remains high-quality and relevant to the topic. Avoid using ChatGPT to create spammy or unhelpful content against Google’s guidelines.