Blog / AI & search /

Bard is out, but not here yet: what it could mean for SEO research

Google opened Bard to the public on 21 March. Well, to some of the public: there’s a waitlist, and it’s only open in the US and the UK.

I’m writing this from Nigeria, so I can’t use it yet. Which, honestly, is a decent position to write from. While half of SEO Twitter is posting screenshots, I’ve had a week to think about what a chatbot from Google could change, and what I want to test the day I get in.

So this isn’t a review. It’s a list of questions.

Why Bard is different from ChatGPT

ChatGPT is a product from a company that doesn’t run a search engine. Bard comes from the company that sends most of the world’s search traffic to websites.

That’s the part that matters for anyone doing SEO. If Google decides a chat answer is good enough for more questions, fewer people click through to the pages those answers are built from. Bard on its own doesn’t do that. It’s a separate product, not part of the search results. But it’s a very public experiment in how Google thinks answers should work.

(It also hasn’t had the smoothest start. In Bard’s launch demo in February, it gave a wrong answer about the James Webb Space Telescope, and that mistake made headlines around the world. A useful reminder that confident isn’t the same as correct.)

What I’ll be testing for SEO research

I already use ChatGPT for parts of my research: brainstorming angles, grouping keywords, explaining things I half understand. These are the questions I’ll put to Bard when I can:

1. Does it know what’s current?

ChatGPT’s knowledge stops at a cut-off date. Bard is meant to draw on information from the web. For SEO that’s huge: algorithm updates, new features and tool changes move fast. If Bard can tell me what changed last month, that’s genuinely useful. If it confidently describes things that never happened, that’s worse than useless.

2. Does it show where answers come from?

This is the one I care about most. An answer without sources is impossible to check, and impossible to cite. Bing’s chat shows links to the pages it used. If Bard doesn’t, it’s hard to trust for research, and it tells us something about how Google values the sites behind the answers.

3. How does it handle local questions?

Most AI tools are noticeably weaker outside the US and UK. I’ll be asking about Nigerian search behaviour, local businesses and questions mixing English and Pidgin. If it can’t handle them, there’s a real gap that local publishers can fill.

4. Is it any good at the boring bits?

Grouping a keyword list by intent, suggesting FAQs for a topic, rewriting a title five ways. If it does these as well as ChatGPT, I’ll use both and compare. If it’s better, I’ll switch for those tasks.

What I think it means, for now

It doesn’t change what good content is. The same advice applies as it did when GPT-4 came out: the value is in what a model can’t know, like your experience, your data and your judgement.

It does make simple, factual content riskier. If your page exists only to answer a one-line question (“what’s the capital of…”), any chatbot can answer it without sending anyone to you. Pages that go deeper, show real examples, or help someone do something are much safer.

It’s worth watching what Google does with search itself. Bard is separate for now. The real question for SEO is whether answers like this start appearing inside the search results, and how they credit the websites they come from.

What I’m doing while I wait

  • Keeping my research workflow in ChatGPT, and noting which tasks I’d want to compare
  • Reading what people in the US and UK are finding, with the healthy suspicion every launch deserves
  • Paying close attention to how Bard handles sources

When Bard reaches Nigeria, I’ll run the same tasks through both and write up what I find.

If you’re in the US or UK and you’ve tried Bard for SEO work, what was it good at? I’d genuinely like to know what to test first.