Goodbye to the “#hashtag”: how AI began the decline of the king symbol of the internet

Because the “hashtag”, that symbol that completely transformed the way content is organized on platforms like Twitter, Instagram or TikTok, seems to be losing its crown. And what was born in 2007 as an ingenious proposal to structure debates on the old internet has become a redundant tool.

However, this progressive disinterest does not respond to a mere aesthetic whim of users, but to a fundamental change in the technological machinery of the networks themselves: behind this phenomenon are the consolidation of semantic Artificial Intelligence algorithms and the rise of social search positioning (Social SEO).

Currently, Recommendation systems no longer need a human being to tell them what an image, video or text is about using a pad, because the technology is capable of interpreting the content autonomously. And, for this reason, the “hashtag” is losing its usefulness.

This decline of the platform does not occur casually or discreetly, it has the direct impetus of the creators and managers of the main platforms on the planet. One of the first to speak clearly about this paradigm shift was the head of Instagram, Adam Mosseri.

Through various virtual meetings with its users, Mosseri bluntly acknowledged that these labels have lost the influence they had in the past to disseminate content, reducing their weight when it comes to gaining visibility.

In their own words, these elements are useful for “categorization, but not for distribution.” Likewise, the manager insisted to the content creators that “turning to them will not substantially alter the reach of their publications.”

In turn, on X (the social network formerly known as Twitter and where this resource achieved its historic popularity), its owner, Elon Musk, has led a technical and visual crusade against the symbol.

In December 2024, Musk addressed his community with a direct request: “Please stop using hashtags. The system no longer needs them and they look ugly.” This stance against labels took a definitive step in June 2025, when the platform banned their use in advertising formats, calling them an “aesthetic nightmare.”

The technical explanation behind this measure is that X’s search engine, supported by Grok (its Artificial Intelligence assistant), already has the ability to organize and understand the context of conversations autonomously, making the presence of the traditional character unnecessary.

For its part, Meta opted for a restrictive strategy with the birth of Threads. Instead of giving free rein to the usual infinite lists of hashtags, the application designed the so-called “topic tags”, limiting their use to just one per publication.

Unlike their predecessors, these labels allow you to add spaces and special characters, integrating more fluidly into the writing of texts. The purpose of this design, as Mosseri himself explained, is “to avoid engagement hacking.”

When less is more.

For more than a decade, the digital marketing sector operated under the premise that the greater the number of tags, the greater the impact of the content. However, data analysis has ended up dismantling this belief.

A study carried out by the analytics firm Socialinsider, after examining more than 75 million Instagram posts, confirmed that the number of tags added to a post does not have a decisive impact on its level of reach or its impressions.

In fact, the report indicates that the highest engagement rate per impression, set at 3.41%, occurred in those contents that used only between three and four hashtags.

This trend has accentuated over time. Industry experts warn that overloading descriptions with generic terms such as #love, #viral or #picoftheday is useless compared to the guidelines of current algorithms.

In addition to reducing the credibility of the profile, this excess visually saturates the user and generates an aesthetic sensation very similar to that of junk mail or spam communications. In fact, both Tiktok and Instagram now allow a maximum of 5 hashtags per post.

Going further, in February 2024, the Chinese platform discreetly suppressed the display of global views accumulated by each label, a key data point that creators and market specialists routinely used to measure the impact of current trends.

By replacing this information with the number of total videos linked to each term, the “hashtag” lost its role as an indisputable barometer to calculate the real audience. Now, the statistics start from the searches.

But why do networks no longer depend on these markers to decrypt the Internet? The explanation lies in the notable advances experienced by Natural Language Processing (NLP) and computer vision.

Today, platforms operate through three levels of semantic evaluation and real-time cataloging. The first is the sound plane, based on automatic transcription.

In a matter of milliseconds, the systems decode the audio of any video or reel. For example, If a gastronomic content creator utters the phrase “the best cheesecake in Madrid” in his recording, the platform processes this message instantly and positions it as the preferred result for those searching for pastries in that area.without having to add tags like #recipes or #madrid.

The second level is the visual, focused on the study of the frames. Image recognition tools evaluate the pixel composition of each scene to extract its context.

If a user shares a recording training in the gym, the system immediately identifies the weights, posture, gestures and the space around them. In addition, the words that appear written on the video are also interpreted automatically using Optical Character Recognition (OCR) technology.

Finally, we find the plane of written language. In this case, the algorithm examines the descriptions and comments, but does not track individual terms preceded by a numeral, but rather deciphers the coherence and semantic structure of complete sentences.

And so, in a digital environment where technology has the ability to observe, listen and assimilate the real meaning of a publication much more efficiently than any manual marker, the pad is relegated to an almost anecdotal role.

Yes, it was a key instrument at the time to train the first neural networks, but today the platforms do without it. And its use responds more to the force of habit of the most veteran users than to any real utility.

The habits of the common user have also undergone a notable change. It is no longer common to browse the Internet by passively clicking on tags or following specific “hashtag” threads. (a function, the latter, that Instagram ended up removing from its users’ home screens due to its low use).

Today, social networks operate very similar to search engines like Google Search: those who browse enter complete questions and natural expressions based on their actual search intent. This transformation in the dynamics of use has consolidated the rise of Social SEO.

Under this scenario, brand profiles and creators who aspire to gain visibility no longer spend their time searching for “magic tags” in the hope of attracting large audiences: instead, they prefer to optimize the texts of their publications in a more organic and fluid way.

As specialist Jade Beason explains, Instagram and other networks now globally evaluate various factors, such as the natural use of keywords in the author’s explanatory text or the information in their biography, to classify content accurately.

Now, The dominant tactic is to prepare detailed and conversational descriptions, logically integrating the terms of interest within the writing.. Thus, the captions of the publications and the creators’ own dialogues are the new guide for the algorithm.

A practice that not only benefits visibility within the internal search engines of the applications themselves, but also contributes to traditional tools such as Google indexing these publications, making it easier for videos from the networks to appear directly in the general search results on the web.

Beyond the technical and commercial implications, the gradual withdrawal of the pad poses a profound cultural transformation in the digital environment. Linguist Adam Aleksic points out that, in its heyday, This symbol embodied a kind of democratic control of information by Internet users themselves.