Digital Marketing: latest News from the World – April Part 1
Welcome to our blog, where innovation meets strategy in the dynamic realm of digital marketing! In this post, we’ll examine the latest developments and emerging trends that are transforming the digital landscape. From recent updates in social media algorithms to AI-powered marketing techniques, these changes offer exciting possibilities for brands and marketers.
Get ready to learn how these advancements can revolutionize your campaigns, broaden your reach, and boost your ROI. We’ll analyze the key developments shaping the future of digital marketing, offering detailed insights to help you stay ahead in this rapidly evolving field. Don’t miss this comprehensive look at the current trends and news that are at the forefront of digital marketing!
Google says goodbye to useless fillers and revives inverted pyramid theory
It seems that someone in Mountain View leaked some journalism books in Google’s knowledge library, because the most recent changes in its “Search Quality Rater Program Guidelines” reference document reflect a firm determination to eliminate the filler content that has become so common in the SEO of some digital media and, instead, prioritize those articles that satisfy users’ curiosity by displaying the most valuable information from the beginning of the text.
In an effort to boost their strategies at any cost, some content creators have adopted this form of writing that maximizes the suspense generated from a title with a “hook” to capture audiences on Google platforms, especially those who aspire to conquer a place in Discover.
With this, they force readers to embark on a long journey through contexts, secondary data and clues that feed their curiosity, but only after going through several paragraphs (and ads integrated in the text), they find an answer to their initial interest, at the end of the article.
In its most recent update of the guidelines that evaluate the quality of content, Google recognizes that these practices could affect the user experience and, therefore, the visibility of this type of content in search results.
What is the inverted pyramid that inspires Google’s new parameters?
Although Google’s update does not refer to it as such, the inverted pyramid technique has been widely used in journalism since the 19th century, when some print media were just being founded and the telegraph began to become popular.
Because of the need to communicate efficiently, journalists of that time began to transmit their news by placing the most relevant information at the beginning of the message, followed by complementary details that were not indispensable for the receivers to understand the central message, in case they were lost due to failures in the telegraphic signals or space limitations on the paper pages.
Essentials first: the key to good SEO in 2025
According to the Google document, the core content of a page should be designed to fulfill its purpose efficiently. This means that the most important elements should appear at the top of the article, preventing readers from having to scroll excessively or navigate through irrelevant text to find the information they are looking for.
“For a note to be considered valuable, the most important information must be at the top. Therefore, we must give the answer we promise in the title at the beginning of the note and try to lead the user to continue reading with good subtitles (H2). The same happens with visual materials: if the article refers to a video, for example, it is important to put it at the top”, explained Manu Parajuá, digital journalist specialized in SEO and audience analysis.
What is filler content and how to avoid it?
Google defines “filler” as low-effort content that does not provide real value and can make a page look longer than it really is, without offering a better user experience. A classic example is when, in a cooking recipe, long introductions about personal experiences are inserted before presenting the list of ingredients and preparation steps.
To avoid penalties and improve the user experience, it is advisable to:
- Go straight to the point: Key information should be in the first paragraphs.
- Use strategic subheadings (H2) to organize information and encourage reading.
- Place multimedia elements (videos, images, graphics) at the top if they are essential to the story.
The January 23, 2025 update of the “Search Quality Rater Program Guidelines” document directly impacts the daily work of those working in digital media, journalists and online content creators.
In short, Google’s recent changes reinforce its vision of providing clear and straightforward answers in the content that appears in the search engine. For this, they use the inverted pyramid theory to recognize the credibility of media and creators in the digital ecosystem.
Artificial intelligence
OpenAI launches its native GPT-4o imager AI
OpenAI has just taken a big leap forward in terms of AI image creation by integrating its new model into GPT-4o. The 4o image generator surpasses the capabilities of the DALL-E family of models, allowing to create more realistic works, with a higher level of detail, to include coherent text or to respect consistency to generate variations, among other useful capabilities.
The developer has already activated the generation of 4o images in the GPT-40 model for the Plus, Pro, Teams and free plans and also within Sora, its AI video generator. OpenAI has informed that it will soon be deployed in the API and in the Enterprise and Edu plans.
Those who wish to continue using DALL-E to create their AI works will be able to access this model through a dedicated DALL-E GPT account.
Capabilities of the 4o imager
OpenAI has developed the 4o imager with a focus on enhancing the usability of the imager. Beyond the attractiveness of the images, they must also serve to communicate, explain or persuade, and to do so they must be consistent, of high quality and clearly arrange the information.
“We train our models in the joint distribution of images and text online, learning not only how images relate to language, but also how they relate to each other. Combined with subsequent intensive training, the resulting model has a surprising visual fluency, capable of generating useful, consistent and contextual images,” the developer explains.
These premises have materialized in a wide range of new capabilities or improved functions. This has not changed the ease of use of the tool, since this model continues to work in a conversational mode. All you need to do is enter the description of the image you want to create or the instructions for generating variations, and ChatGPT-4o will do the rest.
However, you may notice that the AI takes a little longer to generate the images (up to a minute). This is because the model has a longer “thinking” process to create more accurate and detailed artwork.
Improved photorealism and more styles
The 4o image generator model is capable of creating photorealistic works of high quality and detail. This is in contrast to DALL-E’s capabilities, which in this regard were not as advanced as those of other imaging AIs.
Likewise, OpenAI explains that “training with images reflecting a wide variety of image styles allows the model to create or transform images convincingly.”
I have asked the AI to create this image: “A woman climber is sitting on a rock ledge on a mountain. Her partner is just finishing reaching her, climbing below.” I then asked it to do this in a steampunk style, which it recreated to perfection.
This feature has had an unexpected spin-off: the virality of Ghibli-style images, which we told you about extensively in that article. An explosion of adaptations of memes, classic images and photographs to the format of the Japanese animation studio that has been a worldwide success.
Enhanced text inclusion
While DALLE-3 was capable of inserting text into the images it generated, this capability was far from infallible. It often invented new language or misspelled letters. The 4o image generator is much more accurate in this regard, making the inclusion of text effective.
Image variation with consistency
The fact that image generation is now native to GPT-4o itself allows the model to leverage images and text in the context of the chat, promoting consistency. In this way, you can refine your artwork through natural conversations with the tool and create variations while maintaining consistency of the character and context of the artwork.
To test this capability, I asked 4o to generate the image of “a duckling with a flower on its head”. After this, I asked him to “change the style of the image you created to a 3D style”, then to change it to an origami style, and finally to make the duckling crystal and iridescent.
Also, in addition to varying the artistic style, the 4o image generator is also capable of maintaining consistency, adding new elements to the work, applying color changes and even placing our character in new scenarios.
And yes, if you were wondering, it also allows you to generate hyper-realistic variations of images you upload from your device. In this case, I uploaded a photo from Depositphotos and asked it to “Create an image that recreates the image I attached, but the woman is using a video game console and is sitting at the North Pole”.
study
Survey: 42% of people say Google Search is becoming less useful
Google continues to lose ground among younger generations, with 61% of Gen Z and 53% of Millennials using AI tools instead of Google or other traditional search engines. That’s according to a new Vox Media survey.
Why we care. We continue to see early signs that Google search may be starting to slip – from Google’s market share dropping below 90% for the first time since 2015 to other surveys indicating searchers are frustrated with search quality. Things are changing rapidly – from the rise of AI answer engines and generative engine optimization, to the growth of social platforms for discovery (e.g., TikTok, Instagram, Pinterest). Many brands and marketers are still adjusting to these shifts.
By the numbers. When it comes to Google Search:
- 76% said that more than a quarter of Google Search shopping results appear to “be sponsored or promoted.” Only 14% described these sponsored or promoted results as “very helpful.”
- 42% said Google and search engines are becoming less useful.
- 55% said they get information from their community more than online search platforms.
- 52% said they use AI chatbots or alternative platforms (e.g., TikTok) for information instead of Google.
- 66% said the quality of information is deteriorating, making it difficult to find reliable sources.
What they’re saying. According to a Verge slideshow summarizing the survey’s findings:
- “Legacy tech (e.g. Google) and social platforms are rapidly losing ground as trust and authenticity fades, with more people flocking to AI chatbots, niche communities, and platforms like TikTok. This signals a massive shift and opens the door for disruptive entrants that can offer more authentic, trusted experiences.”
About the data. Vox Media partnered with Two Cents Insights to conduct the survey, collecting responses from more than 2,000 U.S. adults in December.
study
Survey: 42% of people say Google Search is becoming less useful
It is no secret that the artificial intelligence race is moving at meteoric speed and that almost every week we see news from the major players. There’s not even a lull on weekends: yesterday, Saturday, Meta launched Llama 4, its new collection of open source artificial intelligence models, with not one but four arriving: Llama 4 Scout, Llama 4 Maverick and Llama 4 Behemoth.
As Meta explains, all of them have been trained on ‘large amounts of unlabeled text, image and video data’ with one goal in mind: that they have extensive visual understanding, which will translate into much more personalized user experiences. And if we take a look at its previous model Llama 3, we will also see that we are looking at much more powerful configurations.
Llama 4 is already here and has some impressive figures
Llama 4 Meta explains that it is its first generation of models with ‘mixture of experts’ (MoE) architecture, more computationally efficient for both training and query resolution. How do they work? Simply put, they divide processing tasks into subtasks and assign them to smaller, more specialized expert models.
Thus, when detailing parameters and experts, we will understand the former as the reasoning capacity or solving skills of the model. For example, Maverick has 400 billion total parameters, but uses only 17 billion active parameters spread across 128 experts.
- Llama 4 Behemoth is an impressive model with nearly two trillion total parameters, 288 billion active parameters and 16 experts, the most powerful to date and one of the most advanced language models. Still in the training phase.
- Llama 4 Maverick has 17 billion active parameters distributed among 128 experts.
- Llama 4 Scout also has 17 billion active parameters distributed, but in this case distributed over 16 experts has 10 million context window tokens.
Llama 4 Maverick will be Meta’s GPT-40
According to Meta’s test battery, Behemoth outperforms GPT-4.5, Claude 3.7 Sonnet and Gemini 2.0 Pro (though not Gemini 2.5 Pro) in tests focused on STEM science-technical areas such as mathematical problem solving.
As for Maverick, it is intended for use as a general assistant or chat (including creative writing), directly rivaling OpenAI’s GPT-4o and Google’s Gemini 2.0, which it outperforms (according to Meta) in certain coding, reasoning, multilingualism, long context, and image interpretation benchmarks. However, it lags behind more recent and powerful models such as Gemini 2.5 Pro, Anthropic’s Claude 3.7 Sonnet, and OpenAI’s GPT-4.5.
Scout’s strengths are document summarization and reasoning over large code bases. As for its enormous context selling, this translates into the ability to parse images and up to millions of words, allowing it to work with extremely long documents.
Meta explains that Scout can run on a single Nvidia H100 GPU, while Maverick requires a DGX system with H100 or equivalent. Of Behemoth he has given no clues, but it will obviously require even more powerful hardware.
Note that none of the LLama 4 models is a reasoning model, as OpenAI’s o1 and o3-mini are. This type of reasoning models check their answers before delivering them, which implies higher reliability at the cost of being slower. Interestingly, Meta has tweaked its Llama 4s so that they refuse to answer conflicting questions less. Explains Zuckerberg’s company, Llama 4 now responds to debated social and political issues that previous versions avoided. In addition, they claim that Llama 4 is “much more balanced” in terms of the topics it directly rejects
Except for Behemoth, which is still unavailable and in training phase, the rest of Llama 4 models can be tested from now on. Thus, you can download the Llama 4 Scout and Llama 4 Maverick models on Llama.com and Hugging Face.
But watch out because also the blue circle of Meta AI from WhatsApp, Messenger, Instagram Direct already uses Llama 4 in 40 countries, so whether you like it or not you will see it in the popular apps. Mind you, for the moment the multimodal functions are limited to the US and only in English.
Is the EU within those 40 countries? No. Those companies, users and users from the European Union can neither use nor distribute the models, according to the European legislation on AI and data protection. In addition, corporations with more than 700 million monthly active users will have to apply for a special license from Meta.
Latest Tech and Digital Marketing Updates
📌Finished the 1st core update of the year. There have been no major changes in the search results with respect to previous core updates. Let’s keep creating quality content…
📌What is that blue circle on WhatsApp? Have you seen it too? Well, it’s Meta’s AI chatbot, which they’ve already included. Let’s put it to the test.
📌New Bizums between Spain, Italy, Portugal and Andorra. From an ecommerce point of view, it opens a way to offer a new payment method.
📌YouTube changes this in Shorts. From April, they will start counting each playback of a Short no matter how long it has been watched, previously you had to reach a minimum.
📌The craze for coffee shops and gourmet spaces. Bigger fashion brands are opening these establishments to diversify revenues, increase dwell time in stores and, being “instagrammable” spaces, increase visibility on rrss and improve image.
📌Speed up Reels on Instagram Are you interested in a video but it’s too long? Press and hold one of the two sides of the screen and accelerate x2, like in TikTok.
📌Google Refine Products takes shape. This new format of product results, allows you to filter from the search engine by size, color, price, stock… all without leaving Google. Available in America, and coming soon to Europe, so get ready.
📌Copilot Search available on Bing. Microsoft debuts Copilot Search, a more conversational search with AI summaries and direct links to sources.
📌Maps gets updated against fake reviews. Google has added Gemini 2.0 to moderate Maps, blocking fake reviews and questionable changes to business profiles. In 2024 they removed 240M fake reviews and blocked 12M fake businesses.
📌Reddit Ads gets a makeover. Launches new features in its ad platform, such as importing campaigns from Meta and 1-click integration with Google Tag Manager.
📌TikTok gets off (narrowly). Trump has given ByteDance 75 more days to sell TikTok’s U.S. operations, avoiding an immediate shutdown. The deadline is mid-June.
📌Heineken “a la Tinder”. They have launched the Bar Dating app, which is like Tinder but for discovering new bars. They seek to support the hotel sector by getting young people to change their usual places.
