Introduction

Every week someone asks the same question: "Can't AI just do our SEO now?"

The honest answer is that AI has genuinely changed how SEO work gets done — and it has quietly made a few parts of it harder, not easier. Search itself has changed too. Google's AI Overviews now appear on a large share of all searches, which means ranking well and getting the click are no longer the same thing.

So here's the practical split: what AI actually does well for your website's SEO, and what it still cannot do no matter which tool you buy.

What AI can do

1. Compress the research work

Keyword clustering, competitor gap analysis, mapping search intent across a topic, grouping hundreds of queries into sensible page structures — this used to take days. AI does the first pass in minutes. You still make the judgement calls, but you make them from a much better starting position.

2. Draft and restructure content

AI is genuinely good at outlines, first drafts, headings, meta descriptions, FAQ blocks, and rewriting a rambling page into something scannable. Google's position is that AI-generated content is not penalised simply for being AI-generated — the policy targets low-quality, manipulative, or mass-produced content made to game rankings rather than help people. Useful, accurate, original content ranks whether a human or a machine typed the first version.

3. Handle technical SEO grunt work

Generating schema markup, spotting broken internal links, writing redirect rules, auditing heading hierarchy, producing alt text for hundreds of images, reviewing crawl logs. This is repetitive, rule-based work — exactly where AI earns its keep.

4. Translate and localise at scale

Multi-language sites used to be prohibitively expensive. AI makes them viable, provided a native speaker reviews the output before publishing.

5. Make content easier for AI search to cite

Clear question-and-answer structure, self-contained sections, plain summaries near the top, proper schema. Google's own guidance is that there are no special technical requirements for AI Overviews or AI Mode — the same SEO fundamentals apply. AI is a good editor for making your pages genuinely clearer, which is the actual mechanism.

What AI cannot do

1. Give you experience it doesn't have

Google's guidance ties its AI features back to its core ranking and quality systems, and industry analysis after the March 2026 core update pointed to Experience being weighted more heavily than before. AI has never installed a boiler, run your shop, or handled your customers' complaints. Your before-and-after photos, your pricing decisions, your "here's what actually went wrong on this job" stories — that's the material AI cannot manufacture, and it's increasingly the material that separates a cited page from an ignored one.

2. Build authority or earn links

Nobody links to a page because it was well prompted. Links, mentions, reviews, and citations come from relationships, original data, genuine expertise, and doing something worth talking about. AI can help you pitch; it cannot make you credible.

3. Verify that it's telling the truth

AI confidently invents statistics, misattributes quotes, and cites studies that don't exist. Publishing that on a business site is a trust problem, not a ranking problem — and on health, legal, or financial topics it's a liability. Every factual claim needs a human check before it goes live.

4. Make mass-produced pages work

Generating 500 near-identical "plumber in [city]" pages is the exact behaviour Google classifies as scaled content abuse. The tooling makes it easy, which is precisely why it's risky. Fewer, better pages beat volume.

5. Fix a slow, broken, or badly structured website

If your site takes eight seconds to load, isn't crawlable, has broken internal linking, or buries the contact details, no amount of AI-written content will rescue it. Technical foundations and design are still design and development problems.

6. Guarantee you traffic in an AI-first search world

Many sites have reported meaningful traffic declines on informational queries where AI summaries now answer the question directly on the results page. The counter-move isn't more AI content — it's targeting queries where people still need to contact someone, buy something, or see proof of work, and building the kind of brand people search for by name.

How to use AI sensibly on your site

• Use it for research, structure, and technical tasks first — the highest return, lowest risk work.

• Never publish an AI draft unedited. Add your own examples, numbers, and opinions.

• Fact-check every statistic, claim, and citation.

• Keep a human voice. Two paragraphs of real experience beat two thousand words of competent filler.

• Measure the right things: branded searches, enquiries, and conversions — not just sessions.

Conclusion

AI is an excellent assistant and a poor strategist. It removes the tedious 70% of SEO work and leaves the 30% that actually decides whether you rank: judgement, expertise, credibility, and a website that works properly.

Use it to go faster on the parts that don't need you. Spend the time you save on the parts that do.