Does AI Content Work for SEO?

AI-written content can rank on Google. Since February 2023 Google has said it judges content on quality rather than on how it was produced, and what it acts against is pages churned out at scale to manipulate rankings, however they were written.

This article covers what Google actually says, where AI content tends to go wrong and how to use AI without putting your rankings at risk.

Google’s position has changed before, and it may change again. In April 2022, before ChatGPT was released, Google’s John Mueller said that text written by AI tools counted as automatically generated content and so went against Google’s guidelines. Google’s guidance on AI-generated content, published on 8 February 2023, takes a more measured line:

Google rewards content that is helpful and written for people, whether a person or AI wrote it. The same guidance says that using AI gives content no special advantage either way. What counts is the value the page adds, which Google ties to E-E-A-T (experience, expertise, authoritativeness and trustworthiness), covered in our guide to E-E-A-T for SEO.

Google has no problem with AI content as such. Its problem is low-quality content.

Content mass-produced to manipulate search rankings counts as spam. Since 5 March 2024 Google has called this scaled content abuse, and its spam policies name generative AI directly: using AI tools to generate many pages without adding value for users is one of the examples they give. A typical case is a website that has AI produce hundreds of pages that tell the reader nothing new. Google says its spam systems look for this kind of content however it was made.

Yes. Google indexes AI-written text, and it can rank for real keywords. Whether it keeps ranking is another matter: we have seen a number of websites whose owners described how quickly AI content built their traffic, only to lose almost all their rankings a few months later.

Google holds a patent on scoring “information gain” (US 11,354,342, granted in June 2022), but it has never confirmed whether or how Search uses it. The idea is simple: if someone reads article A and then article B, how much new information did article B give them? Whatever Google does with it, it is a useful test for your own content.

If your AI text only summarises what is already in the top ten results, its information gain is zero.

That is the biggest trap with AI. Large language models (LLMs) learn from what has already been published, so by default they write the average of it. A page that adds nothing to that average gives Google little reason to rank it above the pages it was built from.

How to add information gain when you write with AI:

• Feed in data nobody else has, such as the results of your own customer survey or internal sales figures
• Take a contrarian view, if you can back it up
• Combine ideas in ways the existing results don’t
• Use your own expertise to decide what stays and what goes

We have seen the same pattern on a number of websites built on AI-generated volume. It is our own observation rather than a published study, and it usually runs in four phases:

Phase 1: Indexing. Google finds the new pages. They contain the right keywords, and traffic starts to rise slightly.

Phase 2: The honeymoon. The pages are tested in the search results. Because they cover lots of long-tail searches, traffic grows quickly.

Phase 3: The plateau. Growth stalls. The pages cover the searches but offer nothing that the pages already ranking don’t.

Phase 4: The update. A core update or a spam update arrives and the traffic falls away.

A strategy built purely on AI volume can work for a while, but once the traffic has gone, it takes a great deal of work, improving pages or removing them, before the domain has a chance to recover.

Ruling out AI altogether means working less efficiently than you could. Used well, it has clear benefits:

• Speed: AI produces a first draft far faster than anyone writing by hand
• Scale: you can publish more with the same team
• Consistency: AI helps keep the basic structure of each piece to the same standard
• A small team goes further: especially in companies without a dedicated content writer

These benefits only hold when someone who knows the subject is steering the tool.

For AI text to help your SEO, it has to meet the reader’s needs. Check every piece against these criteria before it goes live.

The content must be:

• Well written: correct and easy to read
• Informative: it actually answers the reader’s question
• Credible: backed by sources and real expertise
• Original: it adds information gain
• Optimised: structured for both readers and search engines

Warning signs:

• The text lacks depth and nuance, which is common with AI on specialist subjects
• It doesn’t show the expertise and credibility that a niche subject demands

Publishing AI content in bulk is a high-risk SEO tactic with potentially serious consequences. Used for specific, well-defined tasks, AI saves a great deal of time.

Our tips for using AI to produce good SEO content:

• Use it as a sounding board, and keep only what you like
• Ask it to summarise a post you’ve just written
• Ask for very specific information for shorter sections
• Let it draft longer sections that you then edit and improve
• Ask it to suggest common questions (FAQs), and rewrite the ones that fit
• Have it proofread the text and flag errors, which you then check by hand

For more on the writing itself, see our guide to writing blog posts for SEO.

To use AI safely and efficiently, we recommend a process along these lines:

1. Strategy and angle (human): decide what will make this article different. What data or opinion do you have that your competitors don’t?
2. Structure and research (AI): use AI to find content gaps, suggest headings (H2 and H3) and draw up common questions.
3. Draft (AI and human): let AI draft specific sections, but give it your own data to work from.
4. Edit (human): the most important step. Rewrite the introduction, add your own experience and bring the tone into line with your brand.
5. Publish: read the finished article once more and check it sounds like someone who knows the subject.

There are plenty of AI tools that can help you produce content, but a tool is only as good as the person using it.

Widely used AI tools for content and SEO:

• ChatGPT: OpenAI’s general-purpose assistant, useful for brainstorming, structuring and drafting
• Claude: Anthropic’s AI assistant
• Google Gemini: Google’s own AI assistant
• Microsoft Copilot: Microsoft’s AI assistant, also built into Microsoft 365
• Jasper: a platform of AI agents for marketing teams
• Writesonic: began as a writing tool and now focuses on visibility in AI search
• Scalenut: also now focused on AI search, combining AI agents with human specialists
• Hypotenuse AI: built for ecommerce product information, such as product descriptions

Every one of these tools can generate text. None of them knows your subject or your customers the way you do.

• Keep your own tone of voice in everything you publish.
• Show E-E-A-T yourself. AI has no experience of its own to draw on, unless you ask it to make some up, which you shouldn’t.
• Cut any paragraph that adds nothing.
• Cover the topic the way semantic SEO requires, including the related questions and terms a reader expects.
• Web pages aren’t Word documents. Left to itself, ChatGPT hands you something shaped like a report, so give it the structure you want: headings, short paragraphs and lists. How people use your pages can affect your rankings, and a page that is hard to read rarely holds its place for long.
• A few partly AI-written articles are a very different thing from a whole website of AI content with no E-E-A-T signals. Google isn’t short of alternatives, so you have to earn your place in the index.
• On a new website, put extra time into the first articles and let them establish the site before you publish AI-assisted content at volume.

You don’t need any special structured data to appear in Google’s AI Overviews or AI Mode. Google’s guide to its generative AI features says structured data isn’t required for them and that there is no special schema.org markup to add. A page has to meet the same basic test as for ordinary results: it must be indexed and eligible to show with a snippet.

Structured data:

• Article markup is worth having on your articles as good practice, but it isn’t a condition for appearing in AI Overviews
• Google stopped showing How-to rich results in September 2023 and FAQ rich results in May 2026, so neither is worth adding for the sake of a rich result
• Markup doesn’t turn a thin page into a source that gets cited. The content decides that

What Google says you can ignore:

• llms.txt files and other files or markup made just for AI, which Google Search doesn’t use
• Breaking your content into small chunks so that AI can read it
• Rewriting your pages in a special style for AI

What helps instead:

• Content nobody else has: Google calls it non-commodity content, such as first-hand experience rather than a summary of what is already out there
• A clear structure for readers: logical headings, short paragraphs and lists where they help
• Sound technical SEO: pages that can be crawled and indexed, and that work well on every device

Ranking on Google is no longer the only goal. More and more people get their answers from AI tools such as ChatGPT, Perplexity and Microsoft Copilot, and the pages those answers cite are the ones that get seen.

There is no published formula for how these tools choose their sources, so treat the points below as good practice. They are the same habits that make a page useful to a human reader.

How to make your content easy to cite:

• Answer the question first: open each section with a sentence that answers the question in its heading, then explain.
• Use facts that can be checked: name your sources and link to the primary ones, such as the study itself or the platform’s own documentation.
• Show who is behind the text: a named author with real experience of the subject, which is also what Google looks for under E-E-A-T.

Being cited can also cost you clicks: when an AI answer draws on your page, people can get what they need without visiting your website. Good AI search work strikes a balance between being the source that gets cited and giving people a reason to click through to you.

Before you publish a text that AI has helped with, ask these questions:

1. Does it add something new? (Information gain) Is there anything in it that the top search results don’t already say?
2. Is there a person behind it? Is it clear who wrote it, and do they have the experience and expertise to back it up?
3. Is the tone right? Does it sound like your brand, or like a generic encyclopaedia entry?
4. Have the facts been checked? Has someone clicked every link and checked every claim?
5. Would I read this? If I didn’t work here, would I find it genuinely interesting?

If the answer to every question is yes, the text is ready to publish. If not, it needs more work first.

AI and Google will both move on, and what holds today may not hold next year. What won’t change is that the reader has to come away satisfied, however the page was made.

In the end, AI has raised the bar for good content. When anyone can produce text in bulk, the value in the text is what sets it apart. Use AI to speed up research, structure and drafting, and keep the insight that makes a page worth reading, the information gain, human.

If you’d like help with content that earns its place in the search results, with or without AI, see how we work as an SEO agency.

AI content works for SEO as long as someone reviews it and it adds something the search results don’t already have. Here are short answers to the other questions we hear most, each with its source.

Yes. Google’s guidance on using generative AI content (updated in October 2026) calls it critical to fact-check and review all AI-generated content by hand before publishing, including metadata such as page titles, meta descriptions and image alt text. It also suggests telling readers how the content was made, where that makes sense for them.

The review takes a little time. A wrong fact on a service page can cost you the customer.

No. Google’s quality rater guidelines (September 2025 edition) say that using generative AI tools does not on its own determine a page’s level of effort or its quality rating. Main content made with little to no effort, originality or added value gets the lowest rating, whether a person or AI produced it.

The raters’ scores don’t change rankings directly, but they show what Google’s systems are meant to reward: effort, originality and added value.

Not on its own. In November 2025 Google’s John Mueller wrote on Reddit that having a person rewrite AI content doesn’t change a site’s value or make it authentic.

If a text has nothing of its own to offer, new wording won’t fix it. Add what the page is missing: your own data, examples and judgement.

Not reliably. In OpenAI’s own tests, its AI classifier correctly flagged 26% of AI-written texts and wrongly labelled 9% of human-written text as AI. OpenAI withdrew it in July 2023 because of its low accuracy. In a study by Liang and colleagues (Patterns, 2023), seven widely used detectors labelled on average 61% of essays by non-native English writers as AI-generated.

Your time is better spent improving the text than chasing a clean score from a detector.

Human-written content is more common at the very top. Semrush’s study (April 2026) of 42,000 blog posts ranking for 20,000 keywords found that results in position 1 were eight times more likely to be human-written than AI-generated.

Bear in mind that this is a correlation, not proof of cause, and that Semrush classified the texts with an AI detector, the same kind of tool the previous answer warns about. In Zyppy’s survey of 131 SEO practitioners (September 2026), many respondents rated AI-generated content as a negative, but only when it was produced at scale with little added value.

Yes. Google began rolling out AI Mode in the UK on 28 July 2025, and the same principles apply there: the page that gives a clear, accurate answer is the one with a chance of being cited.

If you’d like your content to hold its place both in Google’s results and in AI answers, read how we approach AI search, or start with a free search audit.