Key takeaways
- AI content is not against Google's rules. Pages mass-produced to rank without adding value are scaled content abuse.
- AI-resistant content holds what a model can't summarize: first-hand experience, your own data and a named expert.
- Audit what you already have: keep, improve, merge or remove each page, using Search Console, GA4 and first-hand value.
AI-resistant content is content an AI tool cannot write by summarizing what is already online. It rests on what only you have: your work, your data, your mistakes and opinions you will put your name to.
Using AI does not make content bad for SEO. What Google acts against is pages made in bulk to rank, with nothing new for the reader, however they were produced. So the real question is not whether you used AI. It is whether your page adds anything a model could not.
Is AI-generated content bad for SEO?
No, not by itself. Google said so in its February 2023 guidance about AI-generated content: “Appropriate use of AI or automation is not against our guidelines.” But using automation to generate content “with the primary purpose of manipulating ranking in search results” breaks its spam policies.
That second sentence is the one that matters. Google’s spam policies say scaled content abuse “is when many pages are generated for the primary purpose of manipulating search rankings and not helping users.” The policy applies “no matter how it’s created.” Its first example is “Using generative AI tools or other similar tools to generate many pages without adding value for users”.
Google tightened this policy in March 2024. In an update dated April 26, 2024, it said people would now see “45% less low-quality, unoriginal content in search results” (Google’s announcement).
Google’s guidance on using generative AI content is practical. It says generative AI “can be particularly useful when researching a topic, and to add structure to original content.” It also says: “It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing.” That review covers titles, meta descriptions, structured data and image alt text too.
If you run an online store, Merchant Center has its own rules. The same page says AI-generated product images must carry IPTC metadata, and AI-generated product titles and descriptions must be labeled as AI-generated.
So a page drafted with AI help, checked by a person and built on real experience is fine. Pages generated in bulk to catch every keyword variation are not.
What makes content AI-resistant
Google draws the line itself, in its guide to optimizing for generative AI features. It calls content “commodity” when it is “often based on common knowledge, which could originate from anyone”. Its example is “7 Tips for First-Time Homebuyers”.
The non-commodity example is “Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line”. The first is a list anyone could write. The second is a story only one buyer can tell.
The same guide says: “Don’t just recycle what others on the internet have already said, or could easily be produced by a generative AI model.” That is the clearest definition of AI-resistant content I know.
I made the business case in a June 2026 post on the Adtric blog: “The cost of producing commodity content has fallen to near zero, because anyone can ask an LLM to write it.” Its value has fallen with it. In my experience, commodity content gets answered by AI summaries. Non-commodity content gets cited.
The same problem shows up on service pages, not only blogs. A line like “We deliver quality and value” could sit on any competitor’s site. If a rival could paste your sentence onto their page unchanged, it is not working for you.
A quick commodity test
Before you publish, ask ChatGPT, Gemini or Claude the question your page answers. Read the reply next to your draft. If both say the same things in the same order, your page is commodity content.
Then look for what the AI answer missed. It might be a number from your records, a client case, a mistake you learned from or a firm opinion. Lead with that.
What only you can put on a page
Google’s AI guide asks for “unique expert or experienced takes that go beyond common knowledge and the ordinary.” In practice, that comes from five places:
- First-hand experience. What you did, what you saw and what went wrong, including the steps you actually follow.
- Your own data. Numbers from your Search Console, analytics, sales or project records.
- A named expert. A real author with a real profile, not “Admin” or “Team”.
- Opinions you can defend. A clear position, with your reasons, on the record.
- Original photos and documents. Your work, your team and your premises, not stock images.
The best example from my own work is my website migration guide on the Adtric blog. It covers a small finance bank’s move to a .bank.in domain, which my team at Adtric ran in September 2025. In the year after the move, 805 of the 1,000 addresses that earned the bank the most Google clicks were PDFs. Had we mapped only web pages, four out of five of them would have broken.
The guide also records what went wrong. At the bank’s request, the old homepage showed a message and then forwarded visitors after a few seconds. Google began treating it as a live page again. By mid-November 2025, weekly clicks to the new domain had fallen from about 5,200 to about 1,800. No model could have written either finding, because the numbers existed only in the project’s own data.
Names matter as much as data. I put it bluntly in the same June post: “Pages with named experts win. Pages without lose.” Google’s 2023 guidance says it more gently: consider “accurate author bylines when readers would reasonably expect it”. My guide to showing first-hand experience covers author bios, About pages and dated case studies.
Opinions cost nothing to add. One of mine: “optimize for Google first and the user experience will follow” has it backwards. A model can list both sides of a debate. It cannot take a side and stand behind it.
The same qualities matter in AI answers. My guide to ranking in AI search covers how assistants choose their sources.
How I use AI, and where I draw the line
AI is part of most of my site audits. It makes me faster. It does not make the decisions.
An audit starts with me looking at the site as a visitor would, for a few minutes. Then I use a system I built with Claude. I give it my brief and first findings, and it works through the audit checklists I set up and drafts a report. I verify the findings by hand, especially the critical ones. That check also tells me where to dig deeper and what to cut.
On the bank migration, my team used ChatGPT and Claude to write our URL-matching scripts and to read their output. With about ten days to prepare, we could not do the mechanical work by hand. But as I wrote in the guide: “One rule held throughout: the tools proposed, and a person approved every redirect.”
Content works the same way. Here is the split I recommend:
| Let AI help with | Keep a person on |
|---|---|
| Research and summarizing sources | Checking every fact, number and quote against its source |
| Outlines and structure | Deciding what the page argues and recommends |
| Finding gaps and weak points in a draft | Adding the experience, data and examples |
| Drafting titles, meta descriptions and alt text | Approving everything published under your name |
Google’s reason for the human check is simple. Generative models “don’t retrieve facts, but predict a likely sequence of words based on their training data.” A confident wrong number under your name costs more trust than the time it saved.
Should you tell readers you used AI? Google’s 2023 guidance says: “AI or automation disclosures are useful for content where someone might think ‘How was this created?'” It adds: “Consider adding these when it would be reasonably expected.” If a reader would want to know, tell them.
Why publishing more will not fix it
Publishing more often feels like progress, but frequency is not a ranking requirement. The idea that you must publish three times a week is one of the SEO myths I disagree with. Volume without anything new only adds more commodity pages.
Google’s own self-assessment for helpful content treats this as a warning sign. It asks: “Are you adding a lot of new content or removing a lot of older content primarily because you believe it will help your search rankings overall by somehow making your site seem ‘fresh?’ (No, it won’t)”. It also warns against “changing the date of pages to make them seem fresh when the content has not substantially changed”.
Chasing volume is how a site becomes a digital dustbin. Dozens of short posts overlap, go stale and compete with each other for the same searches. And a blog whose newest post is years old tells visitors the business may have gone quiet.
Neglect is the other half of the problem. When I trained new SEO hires, a common beginner mistake was treating on-page work as a one-time job. Pages need upkeep. The fix is fewer, stronger pages that you keep current.
How to audit your content: keep, improve, merge or remove
A content audit decides what happens to every page you already have. Work through it in this order:
- List every page. Start from your XML sitemap and your CMS’s list of posts and pages.
- Pull Search Console data. In the Performance report, open the Pages tab. Note each page’s clicks, impressions and the queries it shows for. Use the last 12 months, so seasonal pages are not judged on a quiet month.
- Add business results. In GA4, check which landing pages lead to key events, such as inquiry forms and calls.
- Check indexing. Search Console’s Page indexing report lists pages Google left out. “Crawled – currently not indexed” means “The page was crawled by Google but not indexed.” A cluster of these on one topic deserves a closer look. My guide to how Google crawls and indexes pages explains the usual causes.
- Mark overlap. Flag pages that answer the same question or show for the same queries.
- Judge first-hand value. Run the commodity test. Does the page hold anything a model could not write?
Then give every page one of four decisions:
| Decision | Typical signs | What to do |
|---|---|---|
| Keep | Earns clicks or inquiries; facts are current; has first-hand detail | Leave it. Check its facts and links on a set review date. |
| Improve | Impressions but few clicks; facts out of date; useful topic, generic content | Answer the question in the first lines. Add your data, cases and a named author. Change the date only when the content has substantially changed. |
| Merge | Two or more pages answer the same question or share queries | Combine the best parts into the strongest URL. Redirect the others to it with a 301 and update internal links. |
| Remove | No clicks, no inquiries, no useful links and nothing worth saving | Let it return a 404 or 410, or redirect it to a close replacement if one exists. |
Merging pages without losing them
When a page’s URL changes, Google’s documentation on redirects recommends “a permanent server-side redirect whenever possible.” Google uses a permanent redirect as a signal that “the redirect target should be canonical.”
Send each old page to the page that now answers its question, not to the homepage. My migration guide lists homepage redirects among the 12 mistakes we see most often, and they hurt small merges too.
Removing pages, carefully
Google’s page on HTTP status codes says indexed URLs that return a 4xx code “are removed from the index.” A 404 and a 410 are treated the same.
Google’s guidance on core updates is balanced here. It says “Deleting content is a last resort”. It also says “deleting the unhelpful content can help the good content on your site perform better.” So improve or merge first, and remove only what cannot be saved.
This site, as a worked example
In late 2026, I started this audit on ujjwalganesh.com. I am folding thin, overlapping posts into stronger hub articles and redirecting the old URLs with 301s. This article is one of those hubs. Search Console showed that Google had crawled it but left it out of the index. Three other posts covered parts of the same ground: AI content’s pros and cons, content strategy and generic website copy.
I kept their useful ideas. I removed statistics I could not trace to a source. I also removed advice I disagree with, such as publishing weekly to rank. It is too early to report results. Google says changes can take effect in a few days, “but it could take several months”. I will judge this by months of Search Console data, not weeks.
I also followed my own split for this article. AI helped me draft it from my audit notes, case studies and book, and every quote and figure was checked against its source before it went live.
Where to start this month
- Run the commodity test on your five most important pages.
- Add one thing only you have to each: a number, a case, a photo or a firm opinion.
- Put a real name and a short bio on every article.
- Fact-check anything AI drafted before it goes live, including titles and meta descriptions.
- Audit everything else, page by page: keep, improve, merge or remove.
If you want help, my monthly SEO consulting covers this work. It includes content direction (which pages to write and what each must answer) and the page updates themselves. The monthly report covers inquiries from search, and I do not sell a page count.
