To Bard or to not Bard?

Hey, is on the road. Now we have introduced you loads of fascinating and new details about Bard AI. This text shall be helpful to search engine marketing specialists, advertising managers, and anybody who retains abreast of the most recent information. From the article, you’ll study whether or not it is advisable to change your promotion technique and the right way to do it.

Bard is an experimental chatbot from Google. The one distinction from the beloved ChatGPT is that Bard will get data immediately from the web.

Right here is how Bard describes itself:


Bard and Integration with Google

On February 6, 2023, Google started testing Bard. Now, the chatbot can function immediately in search outcomes and generate a prepared reply for customers. There hasn’t been a mass launch but, however snippets with Bard periodically seem for various individuals.

It would look one thing like this:


We assume that there are hyperlinks to sources within the high proper nook. Based mostly on these, Bard prepares a solution to the person’s question. Now there’s a new objective for search engine marketing specialists — to get on this checklist.

That’s why the lead of the search engine marketing staff at and the chief knowledgeable in communication with AI, Artem Demin, armed himself with a considerable amount of espresso and the assistance of search engine marketing specialists to research as many Bard responses as doable.

What does the FAQ from the creators say?

We began with a easy thought: we went to the FAQ and checked out what the creators of Bard say. They’ve ready an in depth reply:


In brief, Bard generates distinctive content material and tries to not duplicate word-for-word data from present sources. If the chatbot does quote textual content from any web page, it offers a hyperlink to it.

Why is the sky blue?

We determined to begin our journey into the world of Bard with such a easy query. The very first thing to concentrate to is the three draft solutions. This means that it’s getting ready completely different variations of the reply to our question.


Now we ask Bard: “Why is the sky blue?” It offers us an in depth reply to this query with hyperlinks to sources and a “Google it” button.


Should you click on the “Google it” button, we get three extra choices for doable questions associated to our question. The outcomes result in Google:


Should you ask Bard considered one of these questions, it can generate a daily reply with out hyperlinks to sources and drafts:


We continued to ask Bard the identical query: “Why is the sky blue?” After some time, humorous conditions began to occur with our question. After asking a few occasions, we acquired this reply:


The primary draft lists a supply hyperlink that leads nowhere:


There isn’t a such supply within the search outcomes, and it doesn’t have a saved copy of the web page within the net archive.


For comparability, we requested Google the identical query. It gives us a “featured snippet” with a highlighted sentence from the NASA Area Place web site:


If we go into this text, we are going to see the reply proper originally of the article with the identical caption “The Quick Reply”:


Google’s reply to the query “Why is the sky blue?” and what Bard generated are completely different. Though Bard, like Google, used data from the NASA Area Place web site to generate the third draft:


The opposite sources that Bard collected for the drafts are additionally within the search outcomes. However they’re much decrease than NASA Area Place:


For instance, Bard considers extra authoritative, because it locations the web site first within the sources. Nevertheless, Google search doesn’t spotlight data from this website.


Reply with out Sources

As we already talked about, generally Bard solutions with out indicating sources of knowledge. There aren’t any underlines and hyperlinks within the generated textual content. Now let’s take a more in-depth have a look at such a solution. We ask Bard the identical query concerning the sky once more.


Then we determined to rephrase the query and ask Google. We needed to match the solutions acquired from Google and Bard for 2 comparable however not similar questions.

The question for Google was the next — “Why the is sky blue?”. The search engine gave the identical NASA Area Place article, however the “The Quick Reply” caption is gone this time. And the supply just isn’t the brief reply however the textual content of the article itself:


Let’s return to Bard’s reply. Curiously, it determined to generate one thing of its personal this time slightly than quoting another person’s textual content. Apparently, the selection of reply format is as much as Bard itself. It decides whether or not to offer a solution with sources or generate a brand new reply.

Artem determined to test if Bard quoted different websites in its reply. We looked for every sentence in Google by including a search operator – “…”. A match was discovered just for one question — “Blue gentle is scattered greater than different colours as a result of it travels as shorter waves” — on the academic platform Numerade.


That is the one website the place Bard’s reply is repeated verbatim. However this quote just isn’t current within the website’s web page code. We checked the Google archive and located a sentence that matches the sentence from Bard’s reply:


We determined to research the place the Numerade website is situated within the search outcomes. It turned out that for the primary question, “Why is the sky blue?” this website just isn’t even within the high 50.


Query No. 1. Why did Bard take this sentence not from the checklist of authoritative sources however from Numerade?

That is the primary query that involves thoughts as a result of this website may be very removed from the highest of Google. Probably, Bard’s new reply simply coincided with the textual content from Numerade.

We determined to ask Bard which internet sites it considers authoritative for the reply to the query, “Why is the sky blue?”. After which evaluate the solutions acquired with the checklist of web sites it can ship.


The websites from the checklist are the primary ones for our question in Google. And NASA Area Place even bought a “featured snippet.”


It additionally turned out that Bard selected the identical assets as sources, the hyperlinks we noticed within the drafts. Let’s analyze the similarity of knowledge on the websites and in Bard’s reply:

  • The phrase “Rayleigh scattering” is discovered solely on essentially the most authoritative website for Bard,;
  • The sentence “Blue gentle is scattered greater than different colours” is present in Bard’s reply and on the website.

In Bard’s solutions, phrases and phrases taken from different sources are sometimes discovered. This was found not solely within the texts that we included within the article but in addition in a thousand others that aren’t talked about within the article. This means that Bard actually takes data from the sources it talks about.

Speculation 1. Websites for the reply are chosen based mostly on their Google search outcomes: the nearer they’re to the highest, the extra possibilities they need to turn out to be a supply for Bard

This speculation emerged throughout work with the question, “Why is the sky blue?”. Bard generated a solution based mostly on web sites that seem on the high of Google. That’s, whoever is first within the checklist of Google search outcomes would be the supply for Bard.

We determined to test the speculation utilizing the question “XRP value prediction”:


Bard compiled an inventory of assets. Subsequent, we determined to test these websites’ positions in Google search outcomes.


Bard’s Sources: Google Search Outcomes:
Pockets Investor







We checked just a few dozen extra queries the place we obtained comparable outcomes. Not one of the sources matched, which suggests the speculation was not confirmed.

Speculation 2. Bard has thematic lists of authoritative websites, based mostly on which it formulates a solution to the person

We got here to this speculation based mostly on the outcomes of the earlier speculation. If Bard doesn’t take websites from the highest of Google as sources, it probably has an inventory of authoritative websites on every matter.

We requested Bard which internet sites it considers authoritative by way of cryptocurrency value predictions:


Bard compiled an inventory of such websites. CoinMarketCap doesn’t have a touchdown web page for the XRP value prediction question. Probably, Bard used the three websites after it to generate a solution to the “XRP value prediction” question.

We determined to check the second speculation on one other question — “FIFA World Cup 2026 prediction.” Bard gave us the next reply:


We additionally requested the sources from which it took the data.


Then we went to Google and repeated the question:


Within the search outcomes, there was solely one of many websites that Bard used — ESPN.

Then we once more requested the chatbot for an inventory of web sites it considers authoritative for the question “FIFA World Cup predictions.”


The checklist matches the sources that Bard used to generate a solution to the “FIFA World Cup 2026 prediction” question. Every part is because it was the final time.

We determined to take a more in-depth have a look at these websites:

The one related supply for answering our question concerning the World Cup is ESPN. The place did the opposite sources come from?

The Solar gives an article analyzing a video from TikTok. On this video, the creator claims to have come from the longer term and is aware of the precise rating and winner of World Cup 2026.


And Polygon and World Soccer Speak don’t even point out the 12 months 2026 of their articles.

What conclusion may be made? Bard makes use of an inventory of authoritative websites for every matter as sources. It doesn’t significantly care concerning the data and whether or not it’s related. The speculation is confirmed.

Speculation 3. Bard determines the authority of web sites based mostly on the positioning’s backlink metrics

We determined to ask how Bard determines whether or not a website is authoritative for a selected area of interest.


The phrases “respected websites,” “well-known,” and “good popularity” counsel the significance of backlinks.


And what did we resolve to do? Appropriate! Conduct a backlink evaluation of the websites that Bard known as authoritative. We began with the websites that had been the premise for producing the reply to the question “FIFA World Cup 2006 prediction.”

Checklist of web sites for backlink evaluation #1. FIFA World Cup:

  • ESPN
  • The Solar
  • Polygon
  • World Soccer Speak
  • BBC Sport
  • Sky Sports activities
  • Aim
  • FourFourTwo
  • Bleacher Report
  • Sports activities Illustrated


We observed that the upper the metrics for Key phrases (Ahrefs), Natural visitors (Ahrefs), Area Score (DR), and Backlinks of a website, the extra authoritative it’s for Bard. The chatbot ranked the websites in its checklist from larger to decrease metrics.

Area age (years), Referring domains, and Pages in Google don’t have an effect on the authority for Bard. Within the desk, you possibly can see the variation of those metrics among the many websites: there isn’t any pattern of enhance or lower.

After that, we determined to research not simply the websites themselves however particular pages:


Hyperlink to the desk (tab “World Cup”)

The extra backlinks a website has, the extra authoritative it’s for Bard. And we got here up with a speculation:

Web site metrics are extra necessary than the metrics of particular pages.

Checklist of web sites for backlink evaluation #2. Crypto value prediction:

Artem urged analyzing the websites that Bard considers authoritative within the topic of cryptocurrency value predictions:

  • CoinMarketCap
  • DigitalCoinPrice
  • PricePrediction
  • Pockets Investor
  • Longforecast
  • Gov Capital
  • TradingBeasts
  • CryptoNewsZ
  • The Financial Instances
  • Forbes

Websites that Bard talked about when requested for “XRP value prediction”:

  • Pockets Investor
  • DigitalCoinPrice
  • PricePrediction


Hyperlink to the desk (tab “Crypto value prediction”)

As we talked about earlier, Bard skipped CoinMarketCap as a result of it didn’t discover appropriate pages on this supply. If we consider the opposite websites, we are able to see that also they are organized by the metrics Key phrases (Ahrefs), Natural visitors (Ahrefs), Area Score (DR), and Backlinks (from larger to decrease).

Along with these metrics, Ref. domains and Pages in Google adopted the identical sample this time. The websites that Bard considers authoritative have larger metrics in these areas.

The Area Age metric varies tremendously from website to website. Subsequently, it isn’t related to authority in Bard’s eyes.

“Why is the sky blue?”: take two

Let’s return to the question concerning the sky. Recall that Bard generated three drafts based mostly on the next sources:


We skipped the second model as a result of Bard didn’t specify the supply.

As within the evaluation above, we fashioned an inventory of 4 websites for evaluation:

  • NASA Area Place

We organized these websites within the order of search outcomes, and here’s what we bought:


Hyperlink to the desk (tab “Sky”)

The clear chief is NASA Area Place, however solely by one metric — Area Score (DR). In all different metrics, the positioning is inferior to

The website considerably surpasses the website in most metrics, however falls behind in DR. This led us to assume that we have to analyze the backlink profile of particular pages extra deeply.

Evaluation of Particular person Web site Pages


Hyperlink to the desk (tab “Sky (supply)”)

The very first thing we found is the connection between the primary place in Google’s search outcomes and Bard’s alternative of sources. In any case, the NASA Area Place web site was within the “featured snippet” in each circumstances.

However how will Bard select an acceptable supply amongst equally respected web sites? This led us to a brand new assumption:

If the sources have equally excessive metrics, Bard takes into consideration the metrics of the precise web page of the positioning the place the data it wants is situated.

Attention-grabbing Questions for Enterprise

We determined to ask Bard about “Pepe coin value.”


This request doesn’t indicate forecasts and assumptions and is predicated on factual information. Nevertheless, Bard gave us an inventory of respected websites much like the sources of knowledge for the response to the question “XRP value prediction.”



This time, Bard generated a big textual content with clickable photos that result in sources. We noticed that CoinMarketCap leads the checklist and is a related supply for this question.

This identical website acquired a “featured snippet” from Google for the question “Pepe coin value”:


Binance is within the third place within the search outcomes. The value proven within the snippet of this website matches the value on and the value in Bard’s reply. Probably, the chatbot used information particularly from Binance.


As we did earlier than, we performed one other evaluation and instantly observed that Binance’s metrics are a lot larger than CoinMarketCap’s.





The decisive elements for Bard had been the acquainted metrics — DR and Backlinks.

Textual content Evaluation of Search Outcomes and Bard’s Responses

We determined to test what responses Bard would generate if the queries are semantically similar however completely different in type. This concept emerged after we acquired completely different solutions by rephrasing the question “Why is the sky blue?” a number of occasions.

  • “Why do I’ve 5 fingers on my hand?”


  • “Why do we’ve 5 fingers?”


The solutions to our queries differ in construction and presentation of the fabric. Probably, this is determined by the queries themselves. The primary question resembles a baby’s question, so the reply is easier. The second question is an grownup’s question, and the reply is way more advanced and detailed. Maybe, Bard can be making an attempt to grasp who precisely is formulating the question.

Should you enter these queries into Google, the websites within the search outcomes and the “featured snippet” shall be completely different:

  • “Why do I’ve 5 fingers on my hand?”


  • “Why do we’ve 5 fingers?”


Subsequent, we analyzed the identical queries in Surfer search engine marketing, the place key phrases from top-ranking websites in Google search outcomes are in contrast. The service highlights the significance of sure parameters in inexperienced, displaying patterns (direct or inverse correlation with a selected issue, for instance, Title size). On this case, the service highlighted the significance of exterior hyperlinks.


The service confirmed a direct correlation with using the phrases “do,” “have,” and “why” within the physique part. These phrases are used extra regularly on websites within the high 10, from 89 to 75 occasions. On websites on the second and subsequent pages of search outcomes, these phrases happen 65-55 occasions.


Evaluation of the load and frequency of phrases utilized in Bard’s response, content material on the web page for the question “Why do we’ve 5 fingers?” and the “featured snippet” didn’t present any correlations.

Pepe Coin Worth


We analyzed the question “Pepe coin value” utilizing Surfer search engine marketing:


Right here we are able to see an inverse correlation between the frequency of the phrases “pepe,” “coin,” “value.” Websites on the primary web page of search outcomes use these phrases much less regularly than websites on the second and subsequent pages.


We additionally observed a direct correlation with the precise match of the question “Pepe coin value” within the physique of the pages. Pages within the high 10 use this phrase as soon as, whereas subsequent pages enhance utilization, by mentioning it 2 or extra occasions.

In its response, Bard didn’t use the direct match; it used the reverse order of phrases — “value of Pepe Coin.”

In Google search outcomes, we additionally observed the point out of the Memetic coin on a number of pages:


These pages point out that the Pepe coin has a Memetic origin. The phrase “Memetic” seems on the pages fairly often. Nevertheless, Bard didn’t use it in any respect. This led us to assume that Bard centered on the primary web page of the coin value, not on its Memetic model.

For comparability, we performed a textual evaluation of content material from the Binance web page and Bard’s response. The primary key phrase, “value Pepe” has virtually the identical proportion of incidence in each texts.

Screenshot 2023 08 07 143732

The outcomes of the textual evaluation of content material from the CoinMarketCap web page differ from Binance and Bard. The primary question, “Pepe value” is talked about thrice. The share of incidence is 0.2% in comparison with 0.9% at Binance and 1% at Bard:



Based mostly on this information, Artem put ahead a speculation:

Bard bases its responses on the evaluation of the primary and most authoritative supply, key phrase density, and the variety of key queries on the web page.

We additionally observed that Bard’s response signifies the place Pepe Coin may be bought. This data is obtainable on the CoinMarketCap web page however is absent on Binance.


We performed a TF-IDF evaluation of Bard’s response and texts from the Binance and CoinMarketCap pages.


Bard (doc 1) and Binance (doc 2) have comparable Time period Depend metrics. Their TFxIDF scores for the phrases “value” and “Pepe” are additionally very shut. Nevertheless, CoinMarketCap (doc 3) has comparable outcomes to Bard by way of the incidence of the phrase “Coin.”

This led us to the next speculation:

Bard took a paragraph with buying areas from the CoinMarketCap web page as a result of this website can be authoritative. If CoinMarketCap emphasizes this data, it signifies that it also needs to be included within the generated response.

Is it Doable to Inflate DR/Backlinks Metrics to Rapidly and Cheaply Get Into Bard’s Checklist of Authoritative Web sites?

For the experiment, we selected the Thefashionisto web site. Why? This web site shortly gained backlinks and considerably elevated its DR from December 2020 to August 2021.

On the time of the evaluation, its metrics had been as follows:



The location has many high quality hyperlinks from trusted assets, for instance, Wikipedia.


In the course of the evaluation, we found that the ranking elevated because of spamming with exterior hyperlinks, and placements had been made on boards and deserted platforms.


Instance from Pingback

Moreover, the positioning has respectable key phrases for which it ranks. For instance, for the question “mens 70s style”:



Nevertheless, after we despatched this question to Bard, it used extra authoritative and bigger websites:


The chosen sources surpassed Thefashionisto in DR and Backlinks metrics. Nevertheless, inflating the metrics is not going to produce the specified impact as a result of Google’s algorithms work higher than exterior hyperlink evaluation companies.

Spamming redirects or direct hyperlinks to your website can deceive companies, equivalent to Ahrefs, and maybe monetize your website. Nevertheless, these strategies is not going to assist turn out to be an authoritative website for Bard. You may also overlook about being included in its supply checklist.


Google has as soon as once more confirmed the significance of the hyperlink profile, not just for search outcomes but in addition for AI.

Create high quality and helpful content material for customers.

Need to be featured in Bard’s response? Enhance and construct a top quality backlink profile. Distinctive and high quality content material will show you how to with this.

Bard’s reply could include direct quotes with supply attribution. The chatbot-generated response is predicated on a normal textual evaluation of phrases and their proportion ratio in texts on the highest 10 web sites.

To extend the possibilities of being included within the checklist of internet sites used to organize Bard’s response:

  1. Develop your hyperlink profile, paying particular consideration to DR and Backlinks.
  2. Don’t artificially inflate your website’s metrics.
  3. When you enter the highest 5 leaders for a selected question, begin conducting common textual content analyses on pages for the quantity and proportion of key phrase inclusions. Use TF-IDF for this. Concentrate on different websites on the high (particularly these larger) and approximate their entry metrics.
  4. Experiment with the TF-IDF metrics of particular person phrases within the key phrase question.

Leave a Comment