@contact1948 Thanks for the feedback. Most NLP that are currently using PHP is not that advanced. Using a better programming language like Python and integration with APIs like Google, IBM is costly. While we can take the short route and provide a simple NLP processing. If we are going to do it, we need to do it with more accuracy. That is why we need to build a few more features I think will help make ours better. We are building out features like knowing which anchor text was used before, how many times it was used, and much more to make sure we minimize suggestions that are not relevant but also based on priorities. Smarter features and functionality takes some time to build :)

@linkilo Yeah, you're right about this being hard to pull off with PHP. But what you could do is use gpt-3.5-turbo-16k with text splitting to extract the main topic(s) of each page, then let it score how closely related the topics are between pages and if they are closely related, process the post data to find anchor text opportunities. This way there's page level relevance as well and not just text passage relevance with lemmatization, and anchor text opportunities would be topics/phrases highly related to the target page topic rather than just keyword/title based text snippets. Something along those lines... If you let users supply their own API key, there's no additional expense for this. The other option would be to run NLP processing on your own servers and let Linkilo pull the results. That's how Linksy does it (crawl the entire site, process NLP, send results back to the plugin). The obvious issue is the ongoing expense. Perhaps including X page crawls for free per month and then using a pay-as-you-go model with the option to turn off NLP processing in settings wouldn't upset lifetime users.

@contact1948 Thanks for the suggestions. There are a few ways to go about this and we are trying to find the best way forward. Some people dont mind using their own API key, some do not, we can't make everyone happy, but our focus is to make sure that the quality of link suggestion and make it user-friendly. We have been working on this for a while now. But as mentioned before, we are building out the infrastructure to make this work better. Things like GSC seeing a keyword drop, or the keyword is increasing, then we want to provide a way to provide our user to build that momentum or recover the drop in rankings, and finding the best NLP to incorporate it. Or if the anchor text has been used too many times compare to the average. We are trying to make this the smartest solution