AI summary:
This article mainly contains the following content, and the estimated reading time is 8~10 minutes
1. Experience feedback from NewBing, Google SGE, and Perplexity search engines
2. Focus on search needs and discuss the respective advantages of AI search and keyword search.
3. External search trends, hybrid search = keyword search + AI search
The development of large language models represented by ChatGPT has, on the one hand, brought about an explosion of AI applications.
On the other hand, the powerful combination of Bing and ChatGPT has also brought the concept of AI search back into the public eye.
Whether it is internal team communication or external social media research, maybe you, like me, have heard or seen the sentence "AI conversational information acquisition methods may replace existing search engines in the future."
The person actually responsible for SEO Colleagues will also be aware of the frequent updates to Google’s algorithm over the past 23 years. So the editor made a bold attempt to only use NewBing and other AI search engines within a month. Can I abandon the existing search engines?
Conclusion: Cannot completely replace search engines
In the following scenarios, current AI search has obvious limitations
1.Multimedia material (pictures, videos) retrieval
Limitation: There is no way to directly present image results in the dialog box
Bing
2.Retrieve category information
Limitation: Penetrate multiple layers of information sources to directly obtain information from web pages
Bing
Perplexity
3.For some specific reports, the exact download address cannot be provided.
Surprise: Scenarios where AI search has advantages
1. It can help me quickly summarize data information and functional information (essentially still relying on the information quality of the web page source). The summary and extraction of AI can basically be used directly without clicking on a specific web page.
2.In terms of shopping selection, it can summarize various online and offline purchasing channels for me based on my account address, and the interaction is more streamlined.
Extension: keyword search VS AI search
The difference between AI search and keyword search
In one sentence, keyword search is based on text rule matching, and AI retrieval uses vector matching technology to find the most relevant results based on semantic similarity.
For example, if you search for "Nike", the keyword search will return you related products of the Nike brand, and the AI search will recognize that "Nike" is a sports brand, so the search results will also include Adidas, Puma, and Anta. and other products that fall within the semantic scope of sports brands.
Therefore, when the user's search intention is only "Nike" brand products, AI search results are likely to give results that do not match the user's intention. Research shows that keyword searches still perform better than vector searches on single phrase queries and exact brand match queries.
Judging from the type of search terms
However, we use search engines not always for precise searches, but also for “exploration and discovery”. If we classify our search input, it can be roughly divided into:
[Clear object] Suitable for keyword search
➔Category search, such as "pen"
➔ Exact search, such as "Apple iPhone14Pro"
➔ Feature search, e.g. “men’s brown shoes”
[Express pain points] Needs, scenarios, pictures, suitable for AI search
➔ Creative searches like “appetizers for gluten-free dinners”
➔ Concept search, e.g. “Running at night, something that makes me visible”
➔ Symptom-related searches such as “alternative treatments for tinnitus”
In the latter three types of search contexts that are biased toward pictures or problem descriptions, it is difficult for traditional keyword searches to match appropriate search results unless extremely complex and flexible keyword rules are set.
Judging from the frequency of search terms
General search is divided into head high frequency search, medium frequency search and long tail search. Considering the value of resource output, keyword searches often only optimize the rules for Fat Head searches.
Extremely low-frequency long-tail searches will be ignored. The search terms in long-tail searches may only appear 1-2 times, but all long-tail search terms added together actually account for a large proportion of searches.
The advantages of AI search have emerged: in the long-tail search scenario, AI can understand the user's semantic intention to find results, without the need for manual algorithm optimization for different long-tail words.
The future will be more of a "hybrid search", that is, using keyword algorithms to solve precise searches for headers and phrases; and using the semantic understanding advantages of AI to solve low-frequency, scene-based long-tail searches.
Remark:
1. The tools used in this experiment are New Bing, Google SGE , Perplexity
2. AI search popular science report How AI search helps boost revenue and cut costs
留言