May 14, 2026 04:12 PM (This post was last modified: Jun 21, 2026 11:05 AM by G Reviews.)
Hello All,
With my experience as a professional python coder, and in implementing everything from stealth browsers, browser automation with anonymizers, WebGL, mobile browser automation, and integrating smart technology into my creations, I can automate almost any browser process.
As I knew eventually I'd be writing this post, I created an HF Smart Poster that can read all the forum categories, select a topic, read that topic, and then intelligently make a smart post regarding everything it has read and a possible solution to the original posters problem, or a general response to whatever the forum topic is. Unlike forum spammers, this does not spam in the traditional sense that one might think. It was not designed that way. It was designed to add healthy and on-topic discussions to forum threads. It responds smartly to every topic and every category. It integrates AI at every level, from selecting which forum categories to read from, which topics to reply to, and balances its smart responses after reading the original posters topic and all the responses from everyone else in that thread and then it intelligently formulates its response and then posts that response. What comes out is either healthy advice, a potential solution to a problem, or a general addition to the topic at hand. Again, this was not built for spamming in negative ways.
And even though I've improved this project to now 100% efficiency after I made the video, I've proven its capabilities and what some simple coding can do.
But the point of this thread is that I can automate just about any repetitive tasks you might have. Everything from CTR on Google, watching youtube videos and playlists, to automating searches and clicking, building profile histories, and basically anything you can do in a browser, I can automate it and integrate AI to perform any writing tasks that you may have. I integrate proxies into everything as well. So what comes out looks naturally realistic, and has the ip address to match it.
And yes, in order to demonstrate this project I did run it against my account here on HF and I was twice sanctioned for using it. But I had to do that to demonstrate some of the automation that I can create. (Anyone can say they can create anything, but I believe proof matters more.)
My other automation projects center around Google, Yelp, Maps, Youtube, CTR, SEO, etc... and I'll be releasing more videos soon after I edit some of them to remove confidential data before the videos are publicly posted.
Suffice to say, if you've got any projects in mind that could use a little automation and some AI magic, let me know.
Interested in discussing your applications for google CTR / SEO benefits / GMB interactions. Although it would take more in-depth analysis / strategy with googles detections, but wondering what your insights are on that. Looking forward to talking.
May 14, 2026 09:53 PM (This post was last modified: May 14, 2026 10:05 PM by G Reviews.)
(May 14, 2026 09:41 PM)thecodershark Wrote: Interested in discussing your applications for google CTR / SEO benefits / GMB interactions. Although it would take more in-depth analysis / strategy with googles detections, but wondering what your insights are on that. Looking forward to talking.
There are only so many things I can talk about regarding google detections, what works, how to get around this and that, etc.. without creating additional competitors and other bot makers. But when working on google properties and google services it is extremely complicated. People who think doing gmb reviews is just writing and dropping a review and that's it are far mistaken. One of the largest misconceptions is that you can use any google account and write a review on any business located anywhere. That alone shows you know nothing about reviews or how they are done. The reviews themselves are not the hard part. What google looks at most is the reviewers account history and trust signals of that account. Then Google also looks at the gmb being reviewed, whether it is in a high-risk category itself, and what its review history has been. Other things that are looked at: is the reviewers google account complete with its bio, picture in the profile, has it done X amount of hours researching, watching videos, the typing speed, typos, the ip address history of the account, whether there have been other recent reviews from the same /24 cidr ip range within the same time frame, etc.. There are thousands of different trust signals that Google looks at. Google is one of the most difficult entities to fool.
I'm extremely knowledgeable on these topics and I'm more than happy to chat with you. Drop me a pm and we'll setup some time to chat.
Edit: Not to knock another man's bread and butter, but there is another reviews seller here in the forums selling 50 cent reviews done through a telegram bot. Those reviews are guaranteed to be deleted with 5-7 days max. Why? Because the accounts he's using are new, which are instantly flagged by google. His second mistake is that the accounts he's using are not local to the businesses he's dropping reviews on. So not only is he penalizing his own accounts, he's also penalizing the gmb's he's dropping those reviews on. This shows me he has zero experience on doing reviews and no regard for the safety of the gmb's he's destroying with his method.
I was updating the amount of mobile metadata tonight that I add to photos that are posted to Google profiles and I decided to make a short video so you could see the automation that goes into not only changing the name of a Google profile, (in the event you buy foreign-made Google accounts because they are often cheaper to buy) but the account picture as well (which now of course has an advanced mobile metadata profile attached to it as you'll see below:)
Now you might ask....why do we do this? Why go through all this trouble? Well....because now we can take stock photos from nearly any source on the planet and make them "real" photos taken by real mobile devices. It is so authentic that it will pass a forensics test for authenticity.
May 21, 2026 03:05 PM (This post was last modified: May 22, 2026 12:06 PM by G Reviews.)
Completed my Youtube video viewer, liker, and auto-commenter with integrated AI tech doing the commenting smartly and intelligently. This will be module # 4 completed for the overall project of warming up Google accounts.
Since is wouldn't be normal to comment and like every video we watch, the system rolls the virtual dice and determines whether or not to comment, like, or subscribe to the channel. 40% of the time we'll like the video, 8% of the time we'll comment on the video, and 3% of the time we'll subscribe to the channel. This keeps our activity realistic.
🔐 loading encrypted image…
2 days ago I made a video for a client showing him how Google accounts are properly warmed up before being deployed for targeted GMB's (Google Map Reviews). Now let me explain what you see in this video:
In the first part we go through the motions of logging into our existing Google account. Then we send a request to AI to get us a list of 30 historical sites, public parks, national monuments, museums, etc which are considered "safe gmb's" to build up the account's review history, improve the trust ranking of the account, and to move up the ranks of "local guide."
So after AI returns the list of 30 safe gmb's to target, we then check on Google Maps for the exact links to those gmb's. We then strip the tracking data that Google returns attached to the url and then add those gmb's to our database. Once we've collected all the urls and stripped all the tracking tags, we then select 5 safe gmb's all within walking distance of each other so we can create a realistic set of clustered reviews. So what Google sees is that we walked the city during the day and went to multiple places and ended up reviewing each place that we went to.
During the warm-up phase we avoid targeting personal businesses, high risk businesses, and strictly stick to first and second rate public attractions. When we start doing the actual reviews, my system will read a majority of al the other reviews left by other people and other bots (because some are just so obvious) and then we'll formulate our own review based on a collective analysis by AI of what others have said.
Since Youtube said my video was too long, they wouldn't host it. So I've self-hosted it here:
Code
https://www.greviews.org/trial32.mp4
🔐 loading encrypted image…
I'm churning through last night's run (because on batches of 20 accounts per run, the average run time per account is 37-45 minutes on youtube for watching 10-12 videos, doing liking, subscribing and commenting) and I'm seeing that the commenting and liking are proceeding along very smoothly. The goal we have to meet here is 1500-1800 minutes of account activity before Google considers the account fully legit.
The majority of that time is youtube naturally, but we add in time with gmail (checking emails, auto-responding, clicking links from our newsletter subscriptions - I'll hit on this later as to why we do very specific newsletter subscriptions), time searching on Google Maps, chrome based general Google searches, and a few other things we add in for good measure.
🔐 loading encrypted image…
We also have many protections built-in to our system. One of those protections is for not wasting time re-watching videos we have already watched. So when the system encounters videos we've already seen (a database is kept on every account of videos already watched) we automatically skip those videos and continue searching until we come across something that we have not watched yet.
🔐 loading encrypted image…
Now on to Fingerprinting:
This is one of the most important topics in the entire system. A unique, stable, and persistent fingerprint across every single account. We spoof everything from the user-agent string to the operating system to our physical hardware, including what type of gpu our system has. We go so far as to include our canvas noise, our audio context noise, device memory, battery power level, and screen size. All the normal things Google expects to collect on every account login.
Tomorrow I will hit on TLS Fingerprinting JA3/JA4 and why this is extremely important when using mobile proxies.
May 22, 2026 03:36 PM (This post was last modified: May 22, 2026 03:49 PM by G Reviews.)
🔐 loading encrypted image…
Not too bad. 199 videos watched. I'm going to add a like counter, a comment counter, and a subscribe counter to the batch later today so I can see all the statistics by tomorrow. This run took around 10-11 hours with each account averaging between 37 and 46 minutes. So we run this daily combined with our other modules to produce the necessary 1500-1800 minutes required for our Google account validity. We had 2 account failures that were login-related, so on the next run we'll simply replace those accounts.
🔐 loading encrypted image…
In one of the above screenshots I showed you a successful video comment. Going back and checking a day later shows us that indeed our comment is still there. Safe, sound, and happy. Proof positive that our system works and works well.
Another thing I'd like to stress the importance of is rotating your mobile ip address between account runs. This step is extremely important because you do not want 2 accounts running on the same exact ip, even if they are mobile ip's. So as part of my system I've integrated an api system that handles these calls automatically.
🔐 loading encrypted image…
Now, because we warm up accounts based on a custom tier 1, 2, and 3 system, we restrict what categories an account can perform reviews on. Newer accounts can only review tier 1 gmb's, and after 15 successful reviews that account can then review tier 1 and tier 2 gmb's. And we may receive a notice after we reviewed the shit out of a particular city already with an army of accounts, which is the yellow [Discovery] warning. That tells the operator that we need to either expand our search area, consider reviewing another city, or expand our search and eligibility requirements to add additional review categories.
So now that we've already successfully reviewed a total of 381 gmb's across the city, it is time to expand operations even further.......
May 23, 2026 10:49 PM (This post was last modified: May 24, 2026 02:44 PM by G Reviews.)
(May 23, 2026 10:23 PM)thecodershark Wrote: Dam this guy knows what the hell he's doing so it seems - this alone has convinced us to move forward with your review service, will be in touch
Thank you for the vote of confidence. As you can see, I study the hell out of every part of this. From account warming with gmail, newsletter subscriptions, youtube (view time, video selection, comments, subscriptions, and likes), chrome searches, third party cookies, and then the review process itself which is the most complicated part of the entire process.
Last but not least, we need to keep track of every completed review we've done. So I've programmed the system to create an excel file (comma delimited) to show the formal name of the GMB, the completed review short link, the name of the poster, the exact gmail account used, and a date and time stamp for when the review was completed.
We don't just keep these records for fun either. We always learn, evaluate, and improve our methods and strategies from what we have completed. The other reason we do this is because clients demand it. Clients always want the short link as evidence that the work was completed. Therefore, the system creates the necessary tracking and book-keeping for the operation. I've got a little more work to do on this because sometimes the short link is hit or miss. I'll resolve that tomorrow. :)
🔐 loading encrypted image…
Further up in the posts I've been making in this thread I was telling you about how when we are doing GMB reviews (Google Maps Reviews) that we need to upload photos with our reviews. I then told you about how we generate our photos with AI and then re-stamp those photos with mobile photo metadata, turning our AI made photo into a believable real photo taken with an authentic and modern mobile phone.
Now I want to show you how that looks in actual practice:
🔐 loading encrypted image…
There are intentional flaws built-in to some of the photos per our rules to AI before generating a photo. How you send the request to AI and against what rules will determine what output you get. In this case we included one of our fingers too close to the lens and it got caught in photo. That's realism. It happens. And then you can see our metadata was re-stamped as an Apple iPhone 14 Pro as the original taker of the photo.
Now let's take a look at our actual public review and what it looks like:
🔐 loading encrypted image…
Since we were reviewing a public park, what is commonly found in a public park? Bike and walking trails. Therefore our generated and stamped photo was right on the money and generated a very realistic object of what one would expect to find during a day at the public park.
Now how do we handle new batches of 2fa secured bulk gmail account purchases from third parties? We need to secure those accounts within minutes of purchasing and receiving them because some shady mother fuckers like to double dip and double sell their inventory. So we must secure our new stock in the quickest manner possible.
So the new accounts from the seller have 2fa security, no recovery email, and since the seller set the passwords we need to dump that shit as quickly as possible and put our own secured password in place that's triple his password strength.
I made a video of course on how we handle this in the most efficient manner possible:
Code
https://www.greviews.org/totp.mp4
Like everything else, completely automated and no human intervention required. We can bang out these changes a thousand times a day.
May 29, 2026 02:29 AM (This post was last modified: May 29, 2026 02:31 AM by G Reviews.)
Today I want to hit on some very important things that most software makers never work on and never implement, causing their bots to be identified, targeted, and eventually destroyed. This revolves around how we build our per-profile setup. Since we don't want Google linking our accounts together, there are literally hundreds of things we need to implement, block, simulate, emulate, and create or hide.
So I wanted a complete picture of exactly what Google sees every time I run an action on any of my Google modules.
🔐 loading encrypted image…
(I captured the technical data during a live run which is why you see the youtube notification at the top of the screenshot.)
The goal here was to make sure that when running hundreds of accounts at once (the warming phase or sending an army of reviewers out) could not be linked to one physical machine. If Google can link hundreds of accounts to a single machine or server, your done before you've even started. So to prevent that, we have to spoof literally hundreds of different signals. And these profile builds change per-account, but are persistent profiles every time we run the accounts. So Google always sees consistent data and a consistent profile.
But once our spoofing is in place, then, and only then, can we proceed to signing in first by password and then by the 2fa challenge. I've designed this to be completely hands-free and no human assistance necessary. Once we're logged in successfully a unique (per run) query is sent upstream to AI to find X amount of popular topics for videos we can watch on youtube. AI returns those values, and evenly distributes those topics to however many accounts we are running per cycle.
I came across a situation where I was spending too much time per day watching rdp's process hundreds of google accounts per day and having to constantly change out the proxy location for each batch. Since accounts have to be warmed up in very specific locations relevant to the client's gmb, that means tons of different locations for tons of different accounts.
In order to speed up the process, and to make our system a bit more autonomous (requiring less human interaction), I deployed 2 api's: one api handles proxy locations based on a pre-determined location per account, and the other api auto-rotates the mobile ip for the accounts in that specific batch that require a specific city.
So if our batch contained 25 accounts needing Austin, TX, 30 accounts need New York City, and 50 accounts need Los Angeles, CA our system can adjust the proxy accordingly and on the fly. This allows us to continually process medium-sized batches 24 hours a day without the need for human intervention, manually changing the proxy info, and letting AI handle the rest of the operations automatically.
Another very important thing that I want to hit on is the value of the timezone that we are sending Google each and every time we login to accounts from different places. If your proxy says you are in a particular city but your timezone says otherwise, Google will pick up on this and immediately flag the account for a data mismatch, which tells Google you are using a proxy.
So today I added a mechanism which first pulls an ip lookup to determine the exact location of the proxy, its latitude and longitude, and then determines the precise timezone we should be located in and adjusts the system clock info that is sent to Google during standard system checks at login time. So this integration was another integral part of keeping our accounts safe.
