The short version: AI is brilliant for the slow, boring parts of SEO, research, briefs, structure, technical audits, and it’s dangerous the moment you let it write the finished article without a human doing the last 20 percent. The businesses winning with AI and SEO right now aren’t publishing more, they’re publishing fewer, sharper pieces built from AI-assisted research and human judgement. Get that order wrong and you’ll do real damage to a domain that took years to build.
What people think “AI for SEO” means, and what it means
Most people I talk to think using AI for SEO means typing a keyword into ChatGPT and getting back a 1,500 word article to paste into WordPress. I get asked about this most weeks, usually by someone who’s already tried it and is confused about why traffic didn’t move, or worse, why it dropped.
Here’s the thing nobody tells you at the start: SEO was never really about words on a page. It’s about trust signals, structure, intent matching, and whether Google’s systems (and increasingly, whether ChatGPT and Perplexity) think your page is the best answer for a real person’s question. AI can help with almost every part of that job except the trust bit, and trust is the bit that takes longest to fake and longest to rebuild once you lose it.
A quick story from a client, because this is where I learned the lesson
Wait, let me tell you what happened with a client of mine, a small accountancy firm based in Kent, ten staff, decent local reputation. In late 2023 their marketing person got excited about AI and published 42 blog posts in six weeks, all AI-generated, lightly edited, targeting things like “how to register a limited company” and “self assessment deadline 2024.”
Traffic went up for about five weeks. Then Google’s helpful content systems caught up with the site and organic traffic fell by roughly 34 percent over the following two months. Nothing was penalised for being “detected” as AI, that’s not how it works and I’ll come back to that. What happened is simpler and worse: the content was thin, generic, and said nothing the top three ranking pages didn’t already say better. Google noticed a site publishing a lot, fast, with no depth, and downgraded trust across the whole domain, not just those pages.
We spent four months pruning 28 of those posts, rewriting 14 with real client examples and the firm’s own numbers (actual VAT thresholds they’d advised on, actual timelines from real filings), and traffic recovered past the original level by month six. The lesson wasn’t “don’t use AI.” The lesson was “AI wrote the skeleton, and nobody put a body on it.”
The uncomfortable bit most SEO guides skip over
Here’s what I won’t sugarcoat: Google has said publicly, more than once, that it does not penalise content for being written by AI. What it penalises is content written for search engines instead of people, which AI is spectacularly good at producing at scale because that’s exactly the gap it’s trained to fill. So the industry sells you “AI detection tools” and “humanising” services as if the machine leaves fingerprints Google can spot. It mostly can’t, and that’s not the real problem.
The real problem is sameness. When ten thousand small businesses all prompt ChatGPT with roughly the same keyword and roughly the same instructions, you get roughly the same article, structured the same way, saying the same three things in the same order. Google’s ranking systems are built, in part, to reward pages that add something the others don’t. Mass-produced AI content, by design, adds nothing. That’s the actual mechanism behind the traffic drops people blame on “AI penalties.” There isn’t a penalty. There’s just no reason to rank you above the twelve other pages that read the same.
This gets harder the moment you step outside English. Translated content usually reads like translated content, and search intent rarely maps one to one from one country to the next, so a phrase that wins in the UK can land nowhere in a smaller language market. That is the part you want someone local for, which is why a specialist Finnish SEO agency exists rather than a translation plugin and a hopeful attitude.
Where AI earns its place in an SEO workflow
None of this means avoid AI. It means put it where it’s strong.
- Keyword and intent research. I use AI to cluster hundreds of keyword variations into groups by searcher intent in minutes, work that used to take a junior researcher a full day in a spreadsheet.
- Content briefs. Feed it the top 8 to 10 ranking pages for a term and ask what questions they answer, what they miss, and what format they use (list, table, video embed). This is where AI saves the most real hours.
- Technical audits. AI is decent at scanning crawl data, flagging duplicate title tags, orphaned pages, and broken internal links across a site with thousands of URLs, work no human wants to do by hand.
- Schema and structured data. Generating FAQ schema, article schema, and product schema markup is a good use of AI, it’s repetitive, rule-based, and low risk to get wrong in a way that damages trust.
- Meta descriptions and title variants. Ask for 15 versions, pick the one that answers the query, discard the other 14.
Where it falls down is anywhere the value comes from a lived example, a real number from your own business, or an opinion someone will disagree with. AI has no experience. It has patterns. Google’s own quality guidelines lean on E-E-A-T, experience, expertise, authority, trust, and the first E is the one AI cannot supply on its own, ever, no matter the prompt.
The seven-step process I run with clients
This is the version I use, adapted for whoever I’m working with that month, whether that’s a solo founder or a marketing team of six:
- 1. Pick the topic from real questions. Pull from actual customer emails, sales call notes, or support tickets, not just a keyword tool. If ChatGPT is already telling people something wrong or thin about your business when they ask it, that’s your topic list right there.
- 2. Run AI-assisted research. Ask an AI tool to summarise what the top ranking pages cover, then ask it what a reader would still be left wondering. That gap is your angle.
- 3. Build a brief, not a draft. Headings, the specific fact or example each section needs, target word count per section. Fifteen minutes, not fifty.
- 4. Draft with AI, but interrupt it. Write the intro and one key example yourself first. Let AI fill structural sections. Never let it write the whole thing start to finish.
- 5. Add one thing AI cannot know. A number from your own data, a client story, a mistake you made. This single step does more for rankings than any keyword tweak.
- 6. Fact-check every claim. AI invents statistics with total confidence. I’ve caught made-up percentages in client drafts three times this year alone.
- 7. Publish slower than you think you should. Two built pages a month will outperform twenty rushed ones within six months, almost every time I’ve tracked it.
If you want someone to build this process for you rather than piecing it together yourself, that’s what an AI implementation coach is for, not writing your content, but building the workflow so your team does steps one through seven consistently instead of skipping straight to step four.
Where AI SEO advice gets dangerously vague
A lot of what’s published on this topic tells you to “use AI for content ideation” and “optimise for E-E-A-T” without ever saying what that looks like on a Tuesday morning with a blank document open. So let me be specific about the traps.
Trap one: asking AI for “SEO-optimised content” directly. This produces keyword-stuffed, oddly repetitive text because the model is pattern-matching on old-school SEO advice from 2015, not on what ranks now. Never use that phrase in a prompt.
Trap two: treating AI-written FAQ sections as a shortcut instead of a summary of things people ask you. If your FAQ doesn’t match the real questions your sales team or support desk hears weekly, it’s decoration, not SEO.
Trap three: ignoring how your site behaves once someone lands on it. You can nail the content and lose the visitor anyway if the site experience is clunky, no chat, no clear next step, dead ends everywhere. This is exactly the gap covered in the guide on AI chatbots for small business websites, because rankings only matter if the visit turns into something once they’ve clicked through.
The AI search shift nobody’s SEO strategy accounts for yet
Here’s something worth sitting with. A growing share of research now happens inside ChatGPT, Perplexity, and Google’s AI overviews before anyone touches a traditional search results page. Being ranked number one on Google in 2026 doesn’t guarantee you’re the answer these tools quote when someone asks “who’s a good AI consultant near me” or “best accountant for a small limited company.”
The pages that get quoted inside AI answers tend to be specific, structured clearly with real numbers and steps (which is part of why I’m building this post the way I am), and mentioned consistently across the web, not just optimised on-site. This overlaps heavily with traditional SEO but isn’t identical to it. If you’re only measuring Google rankings, you’re missing half the picture of how people are finding businesses like yours now.
Where this connects to the rest of your marketing, not just your blog
SEO doesn’t sit in isolation. The same AI-assisted, human-finished approach applies across content generally, and it’s the same principle behind good social media strategy for agencies, use AI to draft and schedule, but keep a real person deciding what’s worth saying and when. It’s also the same discipline behind good customer-facing automation, which is covered in the piece on AI automation for customer service, where the machine handles volume and a human handles judgement calls.
And if you’re running AI note-taking across your team’s client calls to feed research and content briefs, worth knowing where those tools help and where they miss nuance, I wrote about that directly in the piece on AI meeting notes, what they get right and what they miss, because a lot of content teams are now pulling “insights” out of AI meeting summaries that lost the actual point of the conversation somewhere in the transcription.
What I’d tell you to do this week
Pick your three weakest performing pages, the ones getting impressions but almost no clicks in Search Console. Don’t write anything new yet. Run each one through an AI tool with the prompt: “what question is a reader left with after this page, that the top three competitors also don’t answer?” Then write 300 words that close that gap, in your own voice, with one real example or number from your own business. Republish. That’s a smaller, more useful test than any content calendar overhaul, and it tells you within four to six weeks whether this approach works for your specific site.
Frequently asked questions
Will Google penalise my site for using AI to write content?
No, not for the fact that it’s AI. Google has said directly it evaluates content on whether it’s helpful and demonstrates real experience, not on how it was produced. The risk isn’t detection, it’s publishing thin, generic content at speed, which AI makes very easy to do and Google’s systems are specifically built to spot and downrank.
How much of my content should be AI versus human-written?
I’d aim for AI handling research, structure, and drafting of straightforward sections, roughly 60 to 70 percent of the raw words, with a human writing the introduction, at least one real example, and doing a full fact-check pass on every claim before publishing. The human portion is smaller in word count but it’s the part that determines whether the page ranks and holds.
Can AI do proper keyword research on its own?
It can cluster and categorise keywords faster than a human can, and it’s good at spotting intent patterns across large lists. What it can’t do is tell you which keywords your specific audience searches, because it doesn’t have your search console data or your customer conversations. Combine it with your own data, don’t replace your data with it.
Do I still need an SEO strategist if I’m using AI tools?
Yes, for the decisions AI can’t make: which topics matter to your business goals, what your unique angle is, and when a page is finished versus just generated. AI tools speed up the mechanical parts of SEO. They don’t replace judgement about what’s worth publishing in the first place.
Related reading: How to Build a Sales Funnel as a Solopreneur (Without an Agency, a Team, or 40 Hours a Week) and Calendly Marketing Strategy: How They Built a Brand That Wins.
