Why Bing Matters for ChatGPT Product Recommendation
Web search is a big part of AI Product Recommendations (AIPR).
At the point of External Knowledge Search in AIPR, AI systems have a tendency to either use a web search engine or a proprietary method of web search of their own.
In the case of ChatGPT, they have explicitly mentioned Bing and Shopify as their tools in web search.
This makes sense as Microsoft is a major investor in OpenAI.
Simplified Idea of Bing SEO
At a high level, Bing, Google, traditional search engines, and even AI have this motif at the core – to provide a good experience for users when using their search tool.
For this reason, developers behind respective entities will continue to innovate and improve their platforms to retrieve the most relevant, accurate, and reliable results for each query.
This means that search engines like Bing will punish bad actors that show signs of fraud, scam, misinformation, malice, or manipulation.
And while search engines may choose to share tips and guidance on how to understand their ecosystem better, it would be counterintuitive to share all of the trade secrets for how their systems retrieve and rank results.
Intuitive Implication for Bing SEO
When a user enters a question or a keyword-based search, the engine aims to detect the user’s intent first and foremost, then retrieve web pages that may contain the answer to that question or query.
Intuitively, we should check for these elements:
- Does this web page contain the answer to this question or keyword search?
- How do we know that their web page indeed contains the “right answer?” not just what seems like any answer to the question?
- Is this domain entity an authoritative figure on the subject matter?
- If this result was shown, would this user have a good overall experience?
- Broken links, images, and grammar mistakes can hurt the experience.
- Is the content on this web page of high quality?
- Human readability (easy and simple)
The first line of not just Bing SEO, but any SEO, would be to think of these elements before trying to optimized for any technical influences of SEO.
Technical Factors of Bing SEO
The following are a curated set of key points that were technical in nature.
In this case, Bing itself is the Source of Authority in the subject matter. Hence, Narr Theory referred directly to Bing’s publication of Webmaster Guidelines to curate this set of key insights.
Bing, like Google and other traditional search engines, have these 3 components:
- Crawling – how a bot “crawls” through the web to prob “what’s out there”
- Indexing – organizing the crawled information into Bing’s larger knowledge system like a “library of knowledge”
- Ranking – this happens every time a search query is performed – “what web page should be shown first?”
Crawling
Most of the technical components that are relevant for clients reside in the crawling and indexing part. The overarching idea behind the technical is to communicate with machines better of what kind of information is on a webpage, and how to navigate their website.
Sitemap.XML
An XML Sitemap is a machine-readable map of a website designed specifically for web crawlers rather than human visitors. It lists the important pages on a website, allowing search engines and AI crawlers to discover content more efficiently without relying solely on links between pages.
RSS Feeds
An RSS Feed is a machine-readable stream of a website’s published content that allows subscribers and software applications to receive updates whenever new articles are released. Rather than searching the website for new content, an RSS reader can periodically check the feed and notify users or services of newly published posts.
Meta Tags
Meta Tags are hidden pieces of information embedded within a webpage’s HTML that describe the page to search engines, AI systems, browsers, and social media platforms. They typically include information such as the page title, description, author, and instructions for how the page should be indexed. Unlike the visible content on a webpage, meta tags exist primarily to help software understand, categorize, and present the page appropriately in search results and previews.
ARIA tags
ARIA (Accessible Rich Internet Applications) tags are accessibility attributes added to HTML elements to help assistive technologies, such as screen readers, interpret and navigate a webpage. They provide additional context about buttons, menus, forms, dialogs, and other interactive elements that may not be fully understandable from the HTML alone. Unlike XML sitemaps or meta tags, ARIA tags are designed primarily to improve accessibility for people with disabilities rather than to assist search engines or AI crawlers.
Robots.txt
A robots.txt file is a machine-readable set of instructions placed at the root of a website that tells compliant web crawlers which areas of the site they are permitted or discouraged from crawling. Rather than describing the content of a webpage, robots.txt defines the crawler’s access permissions. It can contain rules that apply to all web crawlers or to specific crawlers, such as GPTBot or OAI-SearchBot, making it an important component of a website’s crawl policy.
Structured Data
Structured Data is machine-readable information embedded within a webpage that explicitly identifies the meaning of its content. Using standardized vocabularies such as Schema.org, publishers can label entities such as articles, products, organizations, authors, reviews, events, and frequently asked questions. Unlike HTML, which often requires machines to infer meaning from page structure and text, structured data provides explicit labels that help search engines and AI systems interpret and categorize a webpage more accurately.
Robots.txt
A robots.txt file is a machine-readable set of instructions placed at the root of a website that tells compliant web crawlers which areas of the site they are permitted or discouraged from crawling. Rather than describing the content of a webpage, robots.txt defines the crawler’s access permissions. It can contain rules that apply to all web crawlers or to specific crawlers, such as GPTBot or OAI-SearchBot, making it an important component of a website’s crawl policy. For most web pages, Robots.txt is more about opting out rather than opting in. Crawlers do not have to be “instructed” to crawl – they instead have to be told to stay away if applicable.
Structured Data (Schema.org)
Structured Data is machine-readable information embedded within a webpage that explicitly identifies the meaning of its content. Using standardized vocabularies such as Schema.org, publishers can label entities such as articles, products, organizations, authors, reviews, events, and frequently asked questions. Unlike HTML, which often requires machines to infer meaning from page structure and text, structured data provides explicit labels that help search engines and AI systems interpret and categorize a webpage more accurately.
Aside from these, it would advised for the client to focus more on the human readability, navigation, and overall usefulness.
Search engines, SEO, and AI are all built for and by humans. This is how web pages should be built as well – to benefit humans. The client’s web page should contain relevant, accurate, and credible information and articulated well so that information can be digested easily.
As AI enhances, previously non-technical people will gain a significantly increased ability to manipulate technical aspects of their products and environments.
Simultaneously, search engines and AI systems will likely adapt in a general path aligned with their long-term motifs which is to ensure and protect the overall benefit to humans, often measured by adoption rates.
Indexing
Bing’s Webmaster Guideline, aside from the intuitive and technical factors highlighted above, does not necessarily share much more insights on how it indexes and ranks results. That would be counterintuitive to their prevention of manipulation by bad actors.
Bing does, however, share a tool called Index Now, which is a publicly available to use when a website has gone through any significant changes to help Bing index their content properly.
Narr Theory’s take on how Bing and other search systems index and rank results is that they should be thematically aligned with their long-term motif of showing users the most relevant, accurate, reliable, and quality results.
Indexing is not something to necessarily analyze and strategize too much.
Rather, the higher level themes to keep in mind are that:
- There is a domain of existence for every knowledge, idea, product, or content.
- Search engines and AI are meant to help people find the most relevant, accurate, credible, and easy-to-digest information.
- If the client intends on ranking high for certain queries, they should invest in authority signals, producing high quality content, and using the right technical means to communicate to digital systems what type of content reside in their domain.
Also, search engines and AI systems will increasingly improve at categorizing all web pages and their domains as a certain “type of entity”, in order to better organize their position, nature of existence, and intent in the vast collection of web content and knowledge. (E.g. lifestyle blog, e-commerce store, educational institution, consulting firm, etc.)
Over long-term, it is best to stay true to the client’s nature of existence and post high quality content meant for readers, rather than to try and manipulate the system too much, which in fact could lead to “looking too much like others”, damaging their sense of overall originality.
Ranking
Ranking is perhaps the most intuitive parts to SEO rather than technical.
How would Bing determine if a certain web page should be shown instead of another?
Would would intelligent designers build its algorithm?
What would they prioritize? What would they avoid?
Ranking happens when a user searches something. It will match their intent.
For example, if a user is looking for a specific restaurant in the area, search engines will understand that intent and interpret the keywords in that way, ultimately retrieving the restaurant on Google Maps, their website, and perhaps social media page.
The topic of ranking alone would be a huge research topic of its own.
For simplicity and to keep focus in AI Product Recommendation (AIPR), let’s explore ranking on this topic.
To avoid manipulation, search engines like Bing or AI like ChatGPT are not likely to publicly share too much details about their ranking algorithm.
So, we resort to our intuition – how would intelligent designers program digital systems to answer to a search query?
- Nature & Intent of Search
- For simplicity, let’s keep our search query type to “product search”, “product comparison”, or something related to finding and researching products.
- In this sense, the nature and intent of search is “product research” or “purchase intent”.
- When purchase intent is triggered, however, Bing, ChatGPT, and other digital systems may trigger a different response. For now, we will focus on product research type query.
- Safety Filter
- Is the intent of this search or its potential answer related to human health & safety?
- Relevance
- How relevant is this web page for the intent of search?
- Authority
- Who should answer this question? Which web page, domain, or even type of entity is best suited to answer this?
- Usefulness
- Did previous users with similar search query find this web page’s information useful?
This human intuitive approach to ranking is aligned well with Bing’s own publication of like How Bing delivers search results. For further details, it is encouraged to see Bing’s own publication, or consult with Narr Theory.
Leave a Reply