Google Cloud Vision Pictures Optimization

+400% Visibility Google Cloud Vision Optimization Hack!

We’ve completed a new test for a local Plumber Listing that has seen incredible results in the last 28 days.  For this reason we decided to share our findings during the test, providing you with additional tools and ideas on how to use Google Cloud vision to optimize images.

Table of Content:

The role of images

Find Places through photos

What is Google cloud Vision

Leveraging AL/ML image analysis

Our SEO Test

The role of images

Over the past five years, we have seen a familiar trend in Local SEO wherein Google has increasingly used images in search results (SERPs). Google has expanded the allocated space for images, image size, and count of images, and with advanced image-recognition technology made the images even more relevant to the person searching.

With Google continuously improving its understanding of image content, as well as the entities within, facial sentiments, logos and more, online marketers should now start crafting images for both Google and the audience. Businesses should take advantage of images not just for branding but also to improve rankings and conversion.

Tips to achieve optimum benefits when using images in local SEO:

1. Only use images that add value.

Images can only add value to the content if they are relevant to the topic. While it’s convenient to use stock photos, they tend to be generic and oftentimes not relevant to your concept. Stock photos are also easy to spot and if you have more of them on your website, it will look unoriginal just like many other sites. It’s to your advantage if you use your own original photos and avoid using stock photos as much as possible.

2. Include relevant keywords in the filename.

Google may be smart but it does not see your images when deciding whether it is relevant or not. Its algorithm can only read the details you shared about the photo through the filename and other texts about an image.

Say you published a photo in one of your articles and you saved it using a generic filename such as 2022-06-08_22h06_14.jpg. While it will not hurt your local SEO, it won’t do any good either. You can improve the image for local SEO by including your relevant keyword to the filename.

3. Use photos that perfectly fit the search intent

It’s a good idea to add photos that nicely match the search intent of each targeted keyword. Google cloud vision for this reason is your best ally, as it allows you to analyze the images before uploading them to understand how to best move about them. In addition if you try to search for a topic and just check the Images tab in Google search results, you’ll notice that most images are wide rather than tall. As Google did not lay out clear guidelines for this area.

For a full-width high resolution photo, the recommended dimensions are 2400x1600px. For images within the content, the recommended dimensions are a max-width of 1500px and 1000px vertically. 

4. Optimize the file size without affecting quality.

The file size of your images is directly proportional to the loading speed of your site. So the bigger the size of the photos, the more time they would require to load which can have a negative impact on your SEO. Smaller and lighter images are good for SEO so reduce the size of the image as much as possible without adversely affecting the quality of the photo. For a full-screen background, the optimal file size is 1MB or lower. For small graphics, it should be 300 KB at most.

5. Use overlays wisely.

In addition to the Prohibited Content guidelines, Photos and Videos are subject to the following criteria:

Stylistic adjustments:

Stylistic adjustments (such as applied filters) are acceptable, provided that these stylistic changes are minimal and are not appended elements such as borders, text, collaged images, etc.

Content that makes it difficult for others to understand the environment you’re sharing may be rejected.

Examples are excessively dark or blurry images, significantly rotated compositions, and use of filters that dramatically alter the representation of the place.

Images must be of a sufficient resolution. Exact requirements may vary by photo type and point of upload.

For videos, only upload content that is stable and has good visual quality. Content that is out of focus, shaky, or has bad exposure should not be uploaded. Distorted or corrupted videos are not allowed.

Superimposed text or graphics

Superimposed text or graphics, including promotional content such as logos, are subject to the following requirements:

For 360 photos, superimposed content must be limited to either the zenith or nadir (top or bottom 25% of the equirectangular image), but cannot be present in both.

For traditional digital photos and videos, superimposed content cannot take up more than 10% of the image or video, and must be limited to a single edge.

Superimposed text or graphics must be relevant. 

Distracting superimposed text or graphics are not permitted.

360 Photos

360 photos must wrap 360º without any gaps in the horizon imagery. These images do not have to extend to the zenith and nadir (top to bottom), but between the top and bottom edges of your 360 photo only minor gaps/holes are acceptable.

Minor stitching errors are acceptable but those with significant stitching artifacts may be rejected.

360 photos must be at least 4K (3,840 pixel by 2,160 pixel resolution or greater).

When multiple 360 photos are published to one area, connections between them may be automatically generated. Whether your connections were created manually or automatically, we may adjust, remove, or create new connections — and adjust the position and orientation of your 360 photos — to ensure a realistic, connected viewing experience

When connecting 360 photos, ensure that any links you create are between nearby vantage points. Do not attempt to create connections that will disorient those exploring your connected 360 photos for the first time.

When publishing multiple 360 photos in an area, never use their close pin/dot proximity or the resulting blue line map visualization to write messages, draw symbols, pictures, or otherwise deface the map.

If you’re appending any form of attribution (watermark, authorship information, etc.) to the zenith or nadir of your 360 photo, please note the relevant requirements under the Superimposed text or graphics section above.

Find Places through photos

The amount of space allocated to images in the Local Pack is increasing, in fact, a single business’s Pack images now occupy almost five times as much of the Local Pack.

The section “Find Places through photos” emphasizes more on how Google wishes to focus on reading photographs in order to increase the vocabulary of information describing not only the brands but also the surrounding territories and geographies.

Through photos (such as the photos of Google street maps), Google expands and updates its database of information, creating correlations and suggesting more and more detailed searches and even customized ones based on the user’s profile.

Here are a couple more sections in local search where Google uses images to create context for the search:

In the picture above my search for “Paste Restaurant” in Bangkok returned all available Michelin restaurants under “GMB search results”.

When searching on Google for a hyperlocal area, the results are now strongly tied to a certain number of GMB search results. The search intent is becoming a dominant signal. Pages on your website should reflect the types of questions people would ask when ready to make their decision. Users now use location and situation as a criteria for their decision-making process instead of just plain keywords.And this is why you’ll often see business attributes trigger GMB search results when there were no such things before.

What is Google cloud Vision

Google Cloud Vision is a powerful image analysis tool delivered through an intuitive API. It recognizes and understands a broad set of entities inside of an image.

The Cloud Vision API provides powerful but easy-to-use Image Analytics capabilities. It enables application developers to build the next generation of applications that can see and understand the content of the images. The service enables customers to detect a broad set of entities within an image, from everyday objects to faces and product logos.

The Google Cloud Vision API allows developers to easily integrate vision detection features within applications, including image labeling, face and landmark detection, optical character recognition (OCR), and tagging of explicit content.

Leveraging AL/ML image analysis

Google is increasingly leveraging AL/ML image analysis to improve its understanding of local entities, the classification of the products/services, and the intentions behind each single search.

In some local search queries, Google does not show the cover image of the GMB listings in the 3 packs, but instead, shows the image that best matches the initial keyword.

In addition to the above and based on my research, Google uses image analysis to do the following:

a) Understand how to prioritize search results by creating correlations between images and keywords, and between images and businesses.

b) Determine more details about the authorship of an entity.

c) Create and validate correlations relating to a specific geographical area and other elements across a certain geo-radius.

d) Improve your digital vocabulary by “feeding” on real-time data that by nature changes continuously.e) Assess regional brand penetration.

Our SEO Test

Here is the SEO test conducted by the GMB Crush team:

1) We’ve taken the stock photos off of this listing. With Google AI recognizing stock photos, we recommend focusing only on high quality pictures that can feed the “Google vocabulary” to best describe products and services of the business.

On various tests, the My Business lists with stock photos are the least performing lists since: a) they are certainly not the best choice to represent a business, b) they don’t tell a story, c) they lower the level of credibility of a business profile.

2) We’ve inspected the top 10 images (via Google Images) for “plumbing + location” and “plumber + service” then analyzed them via Google Cloud vision.

During the search and selection of the images, we took note of:

1. The sentiment linked to the top ten images as well as customer satisfaction which was tending towards positive rather than neutral.

2. The type and quantity of objects present in the photos, and making sure they are as close as possible to the original photographs.

3. Which elements remained more in focus in the original photos. When we analyzed the initial photos via Cloud vision, we noticed that focusing or blurring certain elements significantly changed the final result.

3) We hired a freelance photographer to mimic and shoot the photos in the same way as the top 10 pictures as per point 2 above.


At the end of the photo shoot, the images were again analyzed using Cloud vision to be sure they were aligned with those selected in the first phase.

4) We’ve uploaded all the pictures to the listing photo section.

Given the ability of Google to read logos, associate them with related businesses and calculate the level of business penetration on certain locations, we have decided to upload photos in the following manner:

A) All uploaded photos contained the business logo.

B) The photos resulting in the vast majority of cases, photographs with a worker at work have been uploaded to the “at work” section.



C) Given the presence of direct brand searches associated with the owner’s name, some photos of key members have been uploaded to the “team” section, and pages have been dedicated on the website for individual members including a markup schema for the person.

Bonus: For the “exterior” section, we had additional photographs taken to better “feed” Google with additional data relating to nearby places and roads.

5) We’ve created 10 lengthy GMB posts and linked them to each individual service page. All business posts were linked to the service pages and additionally linked to each other with the goal of ​​passing topical relevance.

Bonus: The photographs with the key team members (people of the same family with the same surname as the owner) were used to create posts, mentioning their position and the service areas in which they operate. All GMB posts by team members were linked to the “person” pages and the services mentioned in the person pages were interlinked back to the service pages + area.

Happy Crush!