Google Bard first looks: What we're seeing so far

Google Bard suggests buying links, predicts the next Core Update and refuses to share links to its sources. We're off to a ridiculous start.

Google opened its Bard waitlist today. Hopefully, you get access soon. While you wait, you can get a taste of how Bard works and behaves.

Let’s have a look at what those in the search community – and beyond – are seeing and sharing in early Bard testing.

No links/citations initially. One of our big concerns from the Google Bard preview was the lack of links to sources. Has Google addressed this?

Initially, no. From a tweet by @simonlesser: “No citations, just a link to ‘Google it’. Hilarious answer when asked point blank about its sources.”

Based on this response, Bard apparently had sources for the information it provided – it just didn’t want to share!

However, later it appeared Google Bard started listing some sources for some queries.

As Search Engine Land’s Barry Schwartz tweeted:

Note the addition of the “Sources – Learn more” with three links.

Schwartz also asked Bard why it often doesn’t show sources and citations in its answers. Here’s how Bard responded:

Bard suggests buying links. Even though Google is opposed to link schemes and buying links, Bard seems to be a bit more lenient. “I think it’s a good idea to buy links…” as shared in a tweet thread by @DeanCruddace:

However, after Bard was told that this advice went against Google’s guidelines, Bard admitted its mistake: “You are correct, it is not advisable to buy links.”

Local search. Some interesting implications for local search were highlighted in a tweet by Greg Sterling, former Search Engine Land contributing editor:

• The same query (“handyman in 94118”) produced three different drafts with minor overlap.
• Choosing to “Google it” returned entirely different results.
• None of the Local Pack results appeared in the Bard lists.

Bard says Google uses CTR for ranking. Google warned that “Bard is experimental, and some of the responses may be inaccurate.”

If you want to see an example of that inaccuracy, look no further than @pedrodias asking Bard: “Do you think Google uses CTR as a signal to classify websites?” Bard: “Yes, Google uses CTR as a signal to classify websites.”

Google has repeatedly denied CTR is a ranking signal. Dig deeper:

Speaking of ranking signals that aren’t ranking signals, Google’s John Mueller has said “there is no such thing as LSI keywords”.

So what does Bard say on the subject for ranking? That there is “evidence to suggest that [Google] may do so.” Via @keithgoode:

Bard says next Google core update is March 23, 2023. Google just launched a core update March 15. So is this a hallucination? Or does Bard know something? Via @ryanjones:

SEO metric hallucinations. Thinking of using Bard for keyword research? It apparently can calculate Search Volume, LinkJuice Calories (what?!) and an EAT score (there is no such thing)! Also from @ryanjones:

Garbage in, garbage out? Some of this may be due to AI hallucination. Some also may simply be the fact that Bard was trained on a lot of bad information and misinformation that has been published about SEO.

Always remember: It doesn’t matter whether content comes from AI or a human – bad information is bad information. Always think critically.

Why we care. We’ve been waiting since Feb. 6 to get our hands on Google’s Bard. Like ChatGPT, Google Bard has the potential to be a helpful tool for SEOs. So it’s important to understand the strengths and weaknesses of this generative AI tool.