Blog / AI & search /

AI tools for SEO in 2023: what’s worth trying

Open any SEO tool’s homepage this year and count how many times you see the word “AI”. Writers, outliners, meta description generators, “AI-powered” audits, chat assistants bolted onto keyword tools. Some of it is useful. Some of it is a text box connected to the same model you can already use for free.

It’s been about nine months since ChatGPT launched, and the noise hasn’t calmed down. So here’s where I’ve landed: which kinds of AI tools are worth your time for SEO work right now, what each is genuinely good at, and where I’d keep a human in charge.

The short answer

  • Start with one general chat assistant (ChatGPT, Bard or Bing Chat) and learn to use it well. It covers most of what the specialist tools do.
  • Use AI for the thinking-adjacent work: brainstorming, outlining, grouping keywords, summarising data, drafting meta descriptions, writing small bits of code.
  • Be careful with AI writers that promise finished articles. That’s the part where quality, accuracy and your own experience matter most.
  • Don’t pay for a specialist AI tool until you can name the problem it solves better than the free options.

General chat assistants

ChatGPT

Still the one most people mean when they say “AI”. The free version runs on GPT-3.5. ChatGPT Plus gives access to GPT-4, which is noticeably better at following detailed instructions and reasoning through messier tasks, but it’s a monthly subscription and it’s slower.

For SEO, it’s good at:

  • Brainstorming angles and questions around a topic
  • Turning a messy list of keywords into rough groups (I described my workflow in using ChatGPT for keyword clustering)
  • First drafts of meta descriptions and title options
  • Explaining a technical concept in plain language
  • Writing regex, spreadsheet formulas and small scripts

It’s bad at anything that needs current facts. Its knowledge has a cut-off, and it will confidently make up statistics, sources and URLs.

Plus users also have Code Interpreter, which lets you upload a file (a CSV export from Search Console, for example) and have it analyse the data. That’s genuinely useful for SEO, with one big caution: think carefully about what data you upload.

Bard

Google’s assistant opened up to far more countries in May, including Nigeria, so it’s no longer something I can only write about from the outside. It’s connected to the web, which helps with more recent topics, and it’s free. It tends to be less consistent than GPT-4 for careful, structured tasks, but it’s worth having as a second opinion, especially for anything time-sensitive.

Bing Chat

Built on GPT-4 and connected to Bing search, with sources linked in the answers. Free in Microsoft Edge. The citations are the main draw: you can click through and check where a claim came from, which is more than the others reliably give you.

Claude

Anthropic’s assistant is getting good reviews, particularly for working with long documents. As I write this, it’s only available in the US and UK, so I haven’t been able to use it.

AI inside SEO tools

Many established SEO tools have added AI features: content outlines from the top-ranking pages, AI-written drafts, title suggestions, “content scores” that compare your draft with competitors.

My take:

  • Outline and SERP analysis features can save time, because they combine the AI with real data the tool already has: what ranks, which terms appear, which questions come up. That’s something a general chatbot can’t do on its own.
  • Content scores are useful as a check, not a target. Writing to hit a score tends to produce posts that read like everyone else’s.
  • One-click article writers are where I’m most sceptical. A full article in a minute sounds efficient. It’s also exactly the kind of content that’s easy for everyone to produce, which makes it hard to stand out.

If you already pay for an SEO tool, try the AI features it includes before buying anything new.

AI writing tools

Dedicated AI writers (Jasper, Writesonic and many others) mostly use the same underlying models as ChatGPT, wrapped in templates and workflows. Templates can be helpful if you produce a lot of short-form copy, like product descriptions or ad variations.

For blog content, I’d rather use a general assistant directly, with my own prompts and my own outline, and do the writing and editing myself. Google has said it cares about whether content is helpful, not how it was made, but what Google has actually said also makes clear that mass-produced content made mainly to rank is a problem, whoever or whatever wrote it.

Search Generative Experience

Google’s SGE, the AI answers at the top of search results, is still an experiment in Search Labs and only available in the US. It’s worth watching closely, because it could change how many people click through to websites. But it’s not a tool you can use for SEO work yet, and from here, I can only follow it through other people’s screenshots. My summary of what’s known is in Google’s Search Generative Experience.

Where I’d keep a human in charge

  • Facts and figures. Check every statistic, date and claim. AI tools invent them convincingly.
  • Experience. What you’ve actually seen, tested and learned is the one thing AI can’t supply. It’s also what makes content worth reading.
  • Strategy. Which pages to build, which keywords matter to the business, what to fix first. AI can help you think, not decide.
  • Anything with private data. Read the tool’s data policy before pasting in client information, analytics exports or anything confidential.

A sensible way to start

  1. Pick one assistant and use it daily for a fortnight on small tasks: meta descriptions, outlines, keyword groups.
  2. Keep a note of what saved time and what needed fixing.
  3. Build a handful of prompts that work for you and reuse them.
  4. Only then look at paid specialist tools, with a specific problem in mind.

Which AI tool have you actually kept using after the first week? That’s usually the honest answer to which one is worth it.