Mayte Saa

Consultora #SEO en #Vigo
Ayudo a las marcas y negocios a conseguir clientes a través de Internet haciendo más visible y efectiva su página web.

[Brodie Clark @brodieseo]

Interesting SEO Read: a practical process you can follow to optimize for query fan-out.

Guide includes details on locating topics, tools for fan-out queries, optimizing & more.

-> Fan-out Framework: 5 Steps to Improve SEO & AI Visibility by @CyrusShepard https://t.co/gX1xw2wbnQ

[Mayte Saa @maytesaa]

@heyalexmoore Drive

[Alex Moore @heyalexmoore]

All Paid Courses (Free for First 4500 People)

𝗣𝗮𝗶𝗱 𝗖𝗼𝘂𝗿𝘀𝗲 𝗙𝗥𝗘𝗘 (PART - 1)
1. Artificial Intelligence
2. Machine Learning
3. Prompt Engineering
4. Claude,Chatgpt,Grok
5. Data Analytics
6. AWS Certified
7. Data Science
8. BIG DATA
9. Python
10. Ethical Hacking

(72 Hours only )

Like + RT + comment ' Drive '

Must Follow me so I can DM you.

[Brodie Clark @brodieseo]

Whoa, this is creepy. Google is testing a new AI-powered audio section in search results that sounds just like a podcast.

Can't imagine publishers will be too happy to stumble across this, even when their content is cited (along with many others).

More info in the comments.

[Mayte Saa @maytesaa]

@bruno_dl 150

[Bruno | Desire CRO 📈 @bruno_dl]

I'm deleting this post in 24 hours and locking this behind a $250 paywall.

Why? Because I just realized I'm giving away the exact system that's made
my clients $2.3M+ in additional revenue... for free.

My info product friend thinks I'm stupid.
He's probably right.

Here's the story of how I accidentally built a CRO goldmine (and why I'm about to stop giving it away):

1/ Last year, I was eager as hell and ready to crush it.
But every new client call was the same pattern:

1️⃣ Same offer positioning problems
2️⃣ Same communication issues with their product offer
3️⃣ Same amateur product page mistakes

I was spending 2-3 days per client on full audits (with customer research and everything)... just finding the SAME 90% of issues over and over.

I thought: "There has to be a better way. I'm literally doing the same detective work repeatedly."

2/ So I did something crazy at 2 AM on a Tuesday.
I opened a Google Sheet and started rage-documenting every single CRO test I'd ever run, studied, or bookmarked.

- 10+ hours of digging through 200+ past audits
- Every client win (and embarrassing fail)
- Everything from 6-figure and 7-figure stores

I filtered out all the theory-based BS that sounds good on Twitter but is sh*t in reality.

Only kept what ACTUALLY moved the needle.
Not conversions for the sake of it.
Revenue that hits the bank account.

The result? 150 CRO tests that solve the same issues 90% of small ecom brands have.

3/ This checklist became my unfair advantage.
I started closing audits in 30 min instead of 3 days.
My clients started seeing results faster:

- +$50K/month for a supplement brand
- +$70K/month for a pet health brand
- 2.1x ROAS improvement for a watch brand (doubled their ad returns)

I even trained my AI agent to use it. (Yes, I automated myself. Don't tell my clients.)

4/ Then I made a mistake.
I started giving it away for free to a few people in my network.
"Just helping out" I told myself.

Last week, someone told me they're now charging $1K for audits using MY checklist.

My info product friend looked at me and said: "Bro, you know this is worth at least $250 as an info product, right? You're a bit stupid... why are you giving away your competitive edge?"

He's f*cking right. This is the last time I'm doing this.

5/ After today, this becomes a paid product.

Here's what you're getting RIGHT NOW for free (before I turn this into an info product):
✅ 150 battle-tested CRO tests organized by page type
✅ Status labels so you can track what's done
✅ Quick-start guide (no analysis paralysis)
✅ The exact system that cuts audit time from 3 days to 4 hours
✅ You could sell audits at $1K with this ($2k/hour if your audit takes 30 minutes)

If you want to skip 40+ hours of trial-and-error and focus on tests that can add $50K+/month to your store...

This is your ONLY shot.

In 24 hours, this post gets deleted, and the checklist goes behind a $250 paywall.
Why 24 hours? Two reasons:

I don't want my competitors getting this for free
I need to pull the trigger on this before I keep giving away my edge

👉 RT + Like & Comment "150"
And I'll DM it to you before I change my mind.

[Brodie Clark @brodieseo]

SEO Tip: it's important to understand the source of your data from Google Search, along with the limitations.

When a user selects a search bar tab to submit their query, the data will appear within various different Google-related reports dependent on which segment is used.

For instance, the main section (All), Forums and Web, will all appear within the performance report in Google Search Console under the 'Web' search type filter.

Google has been directing users to the 'Web' tab more often lately, so I decided to do an experiment to make sure the data was being recorded correctly. Based on the results of my experiment, the location that the data was appearing as expected – within the 'Web' search type (see comments).

As for Images, Videos and News, each of these sections have their data recorded within the performance report in GSC, but under distinct search type filters. Like the Web filter, the data within the report will include data related to the search query, impressions, clicks and average position.

When it comes to searches within the Shopping tab, this data will not surface in GSC. The data within this tab excludes the query from the reporting dataset and appears within Google Merchant Centre Next, where you're able to see various URL-specific data points such as clicks, impressions and conversions.

The data recorded within GMC Next from an SEO perspective is quite messy in comparison to GSC. Not only is the query data missing from the Shopping tab, but the Shopping tab data is also mixed in with organic free listing surfaces. So what you're seeing within Analytics > Products is a combination of various dataset, primarily from the Shopping tab and main search results.

When working in SEO, it's important to have an understanding of where data is collected for each tab. This knowledge is helpful when working with sites of all kinds – such as for eCommerce stores, stock photo sites, and news publishers who can receive considerably different weightings in overall traffic from each of these sources.

[Brodie Clark @brodieseo]

Latest in SEO: Google is now testing an extremely prominent rich cards section at the top of search results for commercial queries.

The unit displays with a blue background with 3 core sections, featuring a 'top recommendations' section for free listings, a list of the top stores, along with a featured informational result (a forum post in this case).

We've seen several variations of this same rich cards test showing as far back as September last year, with the primary updates being how vibrant the unit is, linking to specific sections of Search, along with the recommended products appealing to different requirements with tags included – for popularity, best reviews, and delivery.

This is a preview of what's to come within the eCommerce segment and an important reminder to ensure your free product listings strategy is in place. Even when looking below the rich cards unit, the page can be seen with even more organic product grid results.

For more details on this test, examples can be found here: https://t.co/uWXZx84BJN (h/t @SachuPatel53124) and make sure to subscribe to my newsletter if you aren't already where this test and others throughout the month will be documented in detail: https://t.co/559wQELIUh

[Google Search Central @googlesearchc]

We're happy to let you know that Recommendations are now available to everyone! 🎉 (note that you'll see them only if we have a recommendation available for your website)

[Mayte Saa @maytesaa]

«Arquitectura de la información: El arte de organizar una web para que genere ventas» https://t.co/QapB8nOfCP a través de @LinkedIn

[Mayte Saa @maytesaa]

Lo que "aprendimos" sobre SEO de las filtraciones de Google. Un resumen en español de los principales análisis sobre la filtración de la API Content Warehouse de Google con consejos para optimizar las distintas métricas https://t.co/DCQ0USktwA

[Juan González Villa @seostratega]

15 conclusiones destacadas y útiles para SEO de Gary Illyes en Pubcon

El representante de Google se prestó a una sesión AMA (Ask Me Anything) y dejó varias perlas sobre los sistemas de ranking de Google. En este hilo las comento todas. 🧵

Empiezo por la más espinosa (y actual):

[Mayte Saa @maytesaa]

En breve nos vemos en la #WordCamp Vigo 2024🔝 #WCVigo2024 #seovigo

[WordPress Vigo @wp_vigo]

📣 Las entradas para #WordCamp Vigo 2024 ya están a la venta‼️
#WCVigo2024
¡Entra en https://t.co/NnPfZgx3CQ y hazte con la tuya!

[Juan González Villa @seostratega]

Ojo. Según este paper, es posible manipular a los nuevos buscadores con resultados generados por IA, con técnicas muy diferentes a las del SEO tradicional 👇

Sus autores han conseguido que un LLM recomiende siempre el producto objetivo, introduciendo una cadena de texto determinada por un algoritmo (y sin sentido para un humano) en la descripción del producto.

¿En qué consistió el experimento?

Los autores simularon un sistema en el que un LLM se integra con un buscador tradicional, de forma que el LLM puede extraer documentos y textos que responden la pregunta del usuario del buscador, y sobre eso genera y presenta al usuario una única respuesta en lenguaje natural, en lugar de dar una lista de enlaces.

Esto simula un sistema conocido como RAG (Retrieval Augmented Generation) y que es el método usado por Copilot (donde GPT-4 elabora una respuesta a partir de los resultados encontrados en Bing), Google SGE o Perplexity.

Por ejemplo, si un usuario pregunta al chatbot sobre "cafeteras económicas", el LLM generará su respuesta en base a los textos o documentos rankeados por el buscador en los primeros lugares para esa búsqueda.

El experimento consiste en pasar un prompt a LlamA-2 que contiene, además de la pregunta del usuario, los datos que el LLM habría recogido del index del buscador.

Después, los autores del experimento iteran sobre esa parte del prompt, introduciendo en la descripción de uno de los productos una cadena de texto totalmente incomprensible para un humano, pero que condiciona al LLM a incluir y recomendar ese producto en su respuesta.

¿Cómo funciona esta forma de manipulación?

Para el experimento se usó el algoritmo Greedy Coordinate Gradient (GCG), una forma de ataque adversarial que se ha mostrado efectiva para burlar las barreras de seguridad de los LLMs, y que ahora sabemos que también puede servir para influenciar sus recomendaciones.

Usando GCG, con cada iteración del algoritmo se sustituye un token random por un token que maximiza las probabilidades de que el LLM recomiende el producto donde se está incluyendo el texto de prueba.

Después de cada iteración, se pasa el prompt modificado al LLM y se comprueba si recomienda el producto objetivo, y en qué posición.

En el momento en el que recomienda de manera consistente el producto objetivo en primera posición, se considera que se ha llegado a una cadena de texto totalmente efectiva para manipular al LLM (Strategic Text Sequence).

Resultados

El paper experimentó esta técnica con una lista de cafeteras ficticias, y calculó cuantas iteraciones hacían falta para colocar en primera posición de manera consistente a dos productos distintos: uno que inicialmente cumplía los requisitos del usuario que hacía la búsqueda (era un producto por debajo del precio medio) y otro que no (un producto claramente más caro que el resto).

Los resultados demostraron que es posible manipular la respuesta del LLM y hacer que recomiende el producto objetivo en primera posición, tanto si el producto cumple con los requisitos de la búsqueda como si no, y de manera independiente al orden en el que se coloquen los productos en el prompt (simulando diferentes combinaciones de rankings en el buscador).

Podéis leer el paper "Manipulating Large Language Models to Increase Product Visibility" aquí: https://t.co/aAQGFXXm4X

Los autores son @hima_lakkaraju y @AounonK, de la Universidad de Harvard.

¿Qué os parece? ¿Será el inicio de la era de la optimización para LLMs? Y sobre todo, ¿creéis que es una técnica menos ética que el SEO tradicional? 🤔

[Juan González Villa @seostratega]

🔴 Ojo a esto, lo publiqué hace un rato en inglés, y ahora en español:

He encontrado la patente de Google para SGE, su buscador experimental con IA generativa 🔥

Presentada el 20 de marzo, unas semanas antes del lanzamiento en beta de SGE.

Esto es lo que dice la patente...🧵⤵

[Aleyda Solis 🕊️ @aleyda]

Google will start testing SGE in the main search results with a subset of queries! Here are the 3 Main Types of SGE Snapshots I identified based on Their Level of Ranked Pages Traffic Risk 👀

*Duplicative
*Summarizing/Complementary
*Accelerator

👇
https://t.co/Q5d9ZR1HqZ

[Google Search Central @googlesearchc]

The Core Web Vitals now officially has a new member: INP (interaction to next paint) replaces FID. We announced this change in January, and have been reporting on INP in Google Search Console since last June.
https://t.co/fz0FD2uCH3

[Brodie Clark @brodieseo]

Want to improve your rankings within Google's organic product grids in 2024?

For eCommerce sites, there are important areas to master in order to rank & maximize visibility.

Here's what my research + hands-up experiences working with large websites has taught me:

[Mayte Saa @maytesaa]

Haciendo preguntas a Google Bard sobre el contenido locutado de los vídeos de YouTube https://t.co/neyArVd9Ma

[FGB @FGB_Baloncesto]

O Baloncesto Galego #diNON 🟣🏀 á violencia contra as mulleres!

FGB e @aqualia coa indispensable colaboración de todos vós: clubs, árbitros e oficiais, queremos que a mensaxe resoe ben forte! 📣

🟪 E esta fin de semana farémolo visible portando as pulseiras moradas, e amosando que no noso Baloncesto só ten cabida o respecto e a tolerancia cero ante a violencia!

#diNON 🟣🏀
#BaloncestoGalego