SiteSqueeze installs like a standard WordPress plugin. You will need the SiteSqueeze ZIP file and the license key included in your order email.
The same installation process applies to SiteSqueeze Free and SiteSqueeze Pro.
Before you begin
Make sure you have:
Administrator access to your WordPress website
The SiteSqueeze plugin ZIP file
Your SiteSqueeze license key
Your download link and license key are included in the email titled Your SiteSqueeze Free license is ready or in the corresponding SiteSqueeze Pro order email.
If someone else manages your WordPress website, you can forward that email to your website developer or WordPress administrator.
1. Download SiteSqueeze
Open your SiteSqueeze email and select the installation or download link.
Download the SiteSqueeze ZIP file to your computer. Do not unzip the file. WordPress needs the original ZIP file to install the plugin.
If your browser automatically extracts downloaded ZIP files, locate the original ZIP file in your Downloads folder before continuing.
2. Upload SiteSqueeze to WordPress
Sign in to your WordPress administration area.
From the WordPress menu, go to Plugins → Add New Plugin.
Select Upload Plugin near the top of the page.
Select Choose File.
Choose the SiteSqueeze ZIP file you downloaded.
Select Install Now.
WordPress will upload and install SiteSqueeze. This may take several seconds.
3. Activate the plugin
When WordPress confirms that the plugin was installed successfully, select Activate Plugin.
After activation, SiteSqueeze will be added to your WordPress administration menu.
4. Activate your SiteSqueeze license
Installing the ZIP file adds SiteSqueeze to WordPress, but a valid license is required to use the plugin.
Open SiteSqueeze from the WordPress administration menu.
Go to Settings → License.
Copy the license key from your SiteSqueeze email.
Paste the key into the license field.
Select the button to activate or save the license.
Copy the complete license key, including all letters, numbers, and hyphens. Avoid adding spaces before or after it.
5. Confirm that SiteSqueeze is ready
After the license is accepted, confirm that the License screen shows an active license.
The screen should also identify your SiteSqueeze edition:
SiteSqueeze Free: One website, one active Standard Promotion, and one active Placement
SiteSqueeze Pro: The paid SiteSqueeze features and limits associated with your license
You can now begin configuring SiteSqueeze and creating your first promotion.
Installing an updated version
If SiteSqueeze is already installed and you are manually uploading a newer version, WordPress may tell you that the destination folder already exists.
When WordPress displays the current and uploaded versions, confirm that the uploaded version is newer. Then select Replace current with uploaded.
Your existing SiteSqueeze settings, promotions, placements, performance data, and license information should remain in place during a normal update.
As a precaution, maintain a current backup of your website before manually replacing any WordPress plugin.
Troubleshooting
WordPress will not accept the uploaded file
Confirm that you are uploading the original SiteSqueeze ZIP file. Do not upload an extracted folder or one of the files inside it.
The plugin installed, but SiteSqueeze is not available
Go to Plugins → Installed Plugins and confirm that SiteSqueeze is activated.
The license key is not accepted
Copy the entire key from your SiteSqueeze email.
Check for spaces before or after the key.
Confirm that the license is not already active on another website.
Try copying and pasting the key again rather than entering it manually.
You did not receive the license email
Check your spam, promotions, and junk folders for an email from MWD Software.
If you registered for SiteSqueeze Free again using the same email address, SiteSqueeze should resend the existing license rather than create another license.
The installation is taking a long time
Uploading and activating SiteSqueeze normally takes only a short time. If WordPress remains on the same screen for several minutes, refresh the page and check Plugins → Installed Plugins before trying the upload again.
Need help?
If you need help installing or activating SiteSqueeze, reply to your license email or contact us.
Include the email address used to register or purchase SiteSqueeze. You do not need to send your full license key unless support specifically requests it.
I spend a lot of time in Google Analytics, looking at traffic sources, landing pages, and increasingly, how AI platforms are sending visitors my way. But the way most of us use GA4 has a blind spot: it doesn’t show us everything that’s actually requesting files from our server.
That’s the gap HoneyLog is built to fill. It analyzes server and CDN logs to identify AI crawlers, search bots, malicious bots, spoofed traffic, and real human visitors.
HoneyLog Overall Score
Why 3.6/5: Powerful insights and strong value, held back by a rough onboarding experience and limited guidance on what to do with the data.
The easiest way to explain HoneyLog is to say what it isn’t. It’s not a replacement for Google Analytics or Search Console, and it’s not an SEO crawler like Screaming Frog. It isn’t another AI-visibility platform guessing whether ChatGPT mentions your brand for a set of prompts, either. HoneyLog sits underneath all of that, looking directly at requests hitting your server: Googlebot crawling a page, ChatGPT-User pulling a file, a bot hitting robots.txt, a request generating a 404. When I want all of that noise stripped away, I can pull a report for human visitors only.
AppSumo’s listing describes HoneyLog as connecting bot crawling with the human visits that come from the platforms behind those bots. After using it, I’d call that a fairer description than treating it as a GA4 or Screaming Frog “alternative.” They’re not competing tools. They’re different lenses on the same website.
Watching Google actually find my content
This turned into my favorite use case, and it corrected something about my own habits. I don’t routinely submit new content to Search Console. I publish frequently and have basically trusted that my sitemap updates and Google visits often enough to find new pages on its own. I’d never had an easy way to verify that assumption, so I never did.
HoneyLog gave me one. I publish frequently and have trusted that my sitemap updates and Google visits often enough to discover new content. With HoneyLog, I can actually see when Googlebot and other crawlers find my pages, turning something I’d assumed was happening into something I can observe.
The bots showed up faster than I expected
It didn’t take long for HoneyLog to start finding traffic: ChatGPT-User, OAI-SearchBot, PerplexityBot, DuckAssistBot, and Claude-User all showed up in my early data. But HoneyLog goes deeper than “OpenAI visited.” I isolated OAI-SearchBot and found four requests, then added URL and status-code dimensions. All four were for robots.txt. Three returned 200, one returned 301.
That distinction matters. Seeing “OAI-SearchBot: 4 requests” could easily lead me to assume OpenAI was crawling my content. It wasn’t, at least not in that window. HoneyLog gave me enough data to disprove the conclusion I was about to draw from HoneyLog’s own headline number, which is one of the things I like most about it.
A stranger case came from Singapore. GA4 had been showing a surge of Singapore traffic, and HoneyLog showed substantial Singapore activity too. When I filtered HoneyLog’s data to Singapore, every page view in the selected period was classified as Fake Human. That doesn’t prove my GA4 Singapore sessions were bots; the two tools aren’t measuring identical populations, and I don’t want to overstate the connection. But it gave me another angle on something that already looked suspicious, and some of those requests were presenting completely normal-looking browser user agents before HoneyLog flagged them.
That raises the obvious question: how does HoneyLog know? And that’s where my biggest criticism of the product starts.
Good at showing suspicious. Not great at explaining why
HoneyLog’s classifications include Fake Human, Malicious Bot, Suspicious Bot, Generic Bot, Unknown Bot, AI Citation, AI Indexing, and more. That’s a lot of useful categorization, and it’s also a lot to ask an average marketer to understand without help. I looked through HoneyLog’s documentation for an explanation of what triggers a Fake Human classification and couldn’t find one.
If a tool tells me a request came from Googlebot, I understand exactly what that means. If it tells me traffic presenting itself as a normal Chrome browser is actually malicious, I need to understand what signals produced that conclusion before deciding what to do about it. That gap between the sophistication of the data and how well the product teaches you to use it is the biggest weakness I found.
My favorite report might be the simplest one
This is a strange thing to say about a product I bought to investigate bots, but after spending time in Googlebot, ChatGPT-User, Fake Human, Generic Bot, and everything else, the report I keep coming back to is Human Visitors. Just show me the people. In a web full of crawlers, agents, and automated requests, there’s something clarifying about a report built to strip all of that away. GA4 is still more useful for understanding what those visitors actually do once they land, but HoneyLog gives me another way to separate likely human activity from everything else hitting the server.
A world of 404s I normally never see
The Status Codes report produced its own surprise: within roughly my first day of data, I saw more 404 activity than I’m used to seeing in GA4 over much longer stretches. That’s not a knock on GA4, just the difference between client-side analytics and raw server requests. A bot can hit a URL that hasn’t existed in years, guess at a WordPress path that never existed, or follow an outdated link from elsewhere on the web, and none of it becomes a meaningful analytics session.
The harder question is which 404s are worth caring about. HoneyLog’s Raw Logs feature is built for exactly that kind of investigation, letting me narrow a traffic spike by IP, country, bot, and status code. I can already picture building a standing report for 404s from verified crawlers and checking it periodically, which is far more useful than scrolling a giant list of broken requests.
The report builder is the best feature I almost missed
This didn’t jump out at first. It did once I started asking sharper questions. Instead of “Googlebot visited nine times,” I can ask which URLs it requested, what status codes came back, and how long each request took. Instead of “this article got 12 bot requests,” I can ask which bot made them. I ran into that exact case: a URL showed 12 requests that looked interesting until I filtered it and found all 12 were Fake Human. That one filter completely changed what the number meant.
HoneyLog also exports report data to CSV in parts of the app, which made it much easier to dig into Googlebot’s individual requests outside the interface. For anyone willing to dig, there’s a lot here, and that qualifier is doing real work in that sentence.
The average marketer will stall out on “now what?”
This is HoneyLog’s biggest challenge. It hands me information I haven’t had easy access to before, and then leaves me staring at it: my AI bots averaged slower response times than some other crawlers, but is that a problem? I found a Googlebot 404. Should I fix it? HoneyLog flagged something as Fake Human. Should I block it? Googlebot hasn’t crawled today’s article yet. How long is normal before that becomes a concern?
Those are the questions that turn data into action, and right now HoneyLog is much better at supplying the data than answering them. For a technical SEO or developer comfortable digging through logs, that’s probably enough. For the average marketer, HoneyLog needs more interpretation, education, and guidance before the product lives up to what it’s collecting.
Do you even need this if you already have Cloudflare or an MCP?
Some of what HoneyLog shows may already exist in server logs, CDN platforms like Cloudflare, or other technical tools. But that’s not really the point for me. I wasn’t looking at that data before HoneyLog, and I certainly wasn’t using it to answer questions about Googlebot, AI crawlers, suspicious traffic, or 404s. HoneyLog organizes that information in a way that has already helped me find insights and potential actions I wasn’t getting from the rest of my marketing stack.
Bot Conversions is fascinating. I’d rename one metric
HoneyLog attempts to connect crawler activity to actual human referrals, attributing visits by UTM parameters first and referrer data when UTM isn’t available. The idea is compelling: if OpenAI crawls my site 14 times and ChatGPT later sends me a visitor, I want to see both halves of that story.
One thing bothers me, though. HoneyLog calls one metric Conversion Rate. In my data, Microsoft showed two bot requests and five referred human visits, for a 250% conversion rate. The math works. The label doesn’t. Those five people weren’t “converted” from two crawls; it’s a ratio between two different measurements, and I’d rather see it called something like Referral Ratio or Human Visits per Bot Request. The underlying signal is interesting. I just wouldn’t read that number as a conventional conversion rate.
Onboarding needs work
My first night with HoneyLog wasn’t impressive. After connecting it, I could see that HoneyLog was capturing data in Raw Logs, but the rest of the app still wasn’t ready. About half an hour later, I was still seeing: “Hang tight! Waiting for first aggregation, check raw logs for real-time data! This is necessary to see your aggregated insights.” The message told me what was happening, but not how long I should expect to wait. As a new user, I wasn’t sure whether I needed to wait another five minutes, another five hours, or whether something had gone wrong. I eventually contacted support and called it a night.
The next day felt like a different product. Once enough data had accumulated, HoneyLog became dramatically more interesting. That rocky start matters some context: HoneyLog was founded in June 2026, according to its AppSumo profile. This is a young product, and the onboarding experience shows it. I’d like to see a guided first-run flow that explains what’s happening, how long initial processing normally takes, and three or four useful questions a new customer can answer once the data arrives.
What I’d like to see improved
Documentation tops the list. I want a searchable data dictionary that explains every bot classification and metric, Fake Human included, and then goes a step further to suggest what to actually do about what I’m seeing. If Googlebot keeps hitting 404s, tell me why that’s worth investigating. If an AI crawler’s response times look unusually high, give me context for that. If HoneyLog flags malicious fake-human traffic, point me toward next steps, including where a CDN or security service like Cloudflare fits in.
I’d also like more consistent CSV exporting, clearer empty-result states, better sitemap discovery, and more polished field descriptions. Report templates would help a lot here too: New Content Discovery, AI Crawler Errors, Most-Crawled Content, Verified Bot 404s, AI Crawler Response Time. Don’t just hand marketers a report builder. Teach them which reports are worth building.
Tier 1 offers plenty of value for $49
AppSumo’s Tier 1 gives you one website, 150,000 monthly events, six months of data retention, Sitemap Intelligence, and Bot Conversions Analysis for $49. That’s enough capacity for a decent amount of traffic on a single website, while still giving you the features that made HoneyLog valuable during my testing.
For me, there’s plenty of value at that price. I’m already getting insights and identifying potential actions I wasn’t getting from the rest of my marketing stack. If you manage larger or multiple websites and HoneyLog proves equally useful for you, the higher tiers provide more events, sites, users, and retention. But Tier 1 doesn’t feel like a stripped-down entry plan. It gives you enough of HoneyLog to find out whether this kind of server-level visibility is genuinely useful for your website.
So what is HoneyLog actually good for?
I bought HoneyLog expecting to learn more about AI crawlers, and I did. The bigger discovery was realizing how much activity on my own website I normally never see at all. That’s why I’m keeping it: when I have a specific question, I now have somewhere to look. Did Google find yesterday’s article? Is ChatGPT actually crawling my content, or just checking robots.txt? Why did my 404s spike this week? Was that strange traffic real?
Those are questions I’d either struggled to answer before or never thought to ask. HoneyLog makes them observable. The next challenge for the product is making them actionable, and it’s not quite there yet.
I’m already comfortable saying I’m keeping Tier 1. HoneyLog is producing insights, and pointing me toward potential actions, that I’m not getting anywhere else in my current stack. That’s exactly the kind of product AppSumo’s 60-day return window was made for. You can’t fully judge HoneyLog on day one, because its value comes from watching your own traffic accumulate over time. Install it, let it collect real data, start asking questions of that data, and see whether it tells you something you can act on. In my case, it already has.
This content is for educational purposes and reflects my hands-on experience using the product and the information publicly available at the time of writing. Always evaluate tools based on your specific business needs, goals, and workflows before making a decision.
“Marketing’s purpose always is to enhance people’s lives and contribute to the Common Good.” — Philip Kotler
“The customer evolves into an augmented human—someone who relies on digital tools to think, decide, and act.”
“Human-centered experiences are increasingly seen as premium and worth paying more for.”
Book Theme
Marketing 7.0 argues that the next frontier of marketing is not simply better AI, automation, or performance optimization. It is a deeper understanding of the human mind. Kotler, Hermawan Kartajaya, and Iwan Setiawan call this mind-centric marketing: using cognitive science alongside social media, AI, personalization, and immersive technology to understand how customers pay attention, interpret information, form memories, build trust, evaluate value, and make decisions.
The book treats technology as an enabler rather than the destination. As AI makes execution faster and increasingly similar across competitors, human understanding, creativity, authenticity, trust, and memorable experiences become more important sources of differentiation.
Why You Should Read This Book
You should read Marketing 7.0 if you are trying to understand what marketers should still be uniquely good at as AI takes over more research, analysis, content creation, personalization, and optimization. Rather than offering another collection of AI tools and prompts, the book reconnects modern marketing technology with enduring questions about why people notice, believe, remember, choose, and recommend brands.
Its strongest value is strategic. It gives marketers frameworks for thinking about AI-driven commoditization, digitally augmented customers, cognitive biases, storytelling, trust, perceived value, selling, and customer experience. It is less useful if you want a tactical channel playbook or step-by-step instructions for using specific AI platforms.
Key Ideas and Arguments Presented
The human mind is marketing’s next frontier. Marketing 7.0 shifts attention from optimizing technology to understanding how customers process information, form memories, and make decisions.
Customers are becoming “augmented humans.” Digital tools and algorithms increasingly help people search, filter, compare, decide, and act, which changes how marketers must earn attention and trust.
AI efficiency can create sameness. When competitors use similar platforms, algorithms, data, and optimization techniques, performance marketing can push brands toward increasingly homogeneous output.
Technology and humanity should complement each other. AI excels at speed, scale, pattern recognition, and automation. Humans remain essential for creativity, empathy, context, culture, meaning, judgment, and authentic connection.
Three gateways influence the mind. The authors organize modern engagement around social, personal, and experiential gateways. People are influenced by others, by personal relevance and identity, and by experiences that create memory and emotion.
Cognitive mapping can produce deeper customer insight. Marketers should study how customers actually pay attention, respond to social signals, evaluate rewards and sacrifices, and move toward decisions rather than relying only on stated preferences.
Storytelling works because the mind organizes information through narrative. Effective brand stories create meaning and emotional resonance rather than merely transmitting product claims.
Value is mentally constructed. Customers weigh perceived benefits against sacrifices, and cognitive biases influence that calculation. Product, price, place, and promotion therefore affect not just economic value but perceived value.
Trust is central to selling. Sales communication should match how people process information and uncertainty, reducing cognitive friction and building enough confidence to act.
Customer experience should create memorable moments. Brands should map the journey, identify high-impact moments, deliberately create positive surprises, engage multiple senses where appropriate, empower employees, and make worthwhile experiences easy to share.
Book Outline
Part 1: The Path to the Human Mind
Chapter 1: The Path to Marketing 7.0: Why Social Media, Artificial Intelligence, and Immersive Tech Unlock the Human Mind
Chapter 2: The Threats of Artificial General Intelligence: Why Minds and Machines Must Coexist for Better Marketing
Chapter 3: The Pitfalls of Marketing in the Digital World: Why Performance and AI Obsessions Kill Authenticity
Chapter 4: The Challenges of Engaging Augmented Humans: Why Filtering, Fragmentation, and Frugality Define the New Market
Part 2: The Keys to Unlocking the Human Mind
Chapter 5: The Mind-Centric Marketing: A Guide for Thinking Marketers in the Age of AI
Chapter 6: The Cognitive Mapping: Observing the Human Mind in Action to Extract Deep Customer Insights
Part 3: Essential Marketing Stimuli for Engaging the Human Mind
Chapter 7: Brand Storytelling: Leveraging Three Storytelling Paths into the Customer’s Mind
Chapter 8: Value Proposition: Tapping into Human Biases to Guide Mental Trade-Offs
Chapter 9: Selling Approach: Building Trust to Convert with Confidence
Chapter 10: Customer Experience: Designing Memorable Moments to Inspire Advocacy
Appendix: Key Elements of Marketing: The Relevant Frameworks Behind the Core Models of Marketing 7.0
Key Takeaways
More powerful marketing technology increases rather than eliminates the need to understand people.
AI should augment human marketers, not become a substitute for strategy, judgment, creativity, and empathy.
Optimization without differentiation can make brands efficient but forgettable.
Attention, trust, perceived value, memory, and social influence deserve as much attention as clicks and conversions.
Marketers should observe actual customer behavior and decision-making, not depend entirely on surveys, personas, or what customers say they want.
Brand storytelling, value propositions, sales conversations, and customer experiences can all be improved by understanding cognitive processes and biases.
Customer experience is not merely service delivery. Memorable moments can become a strategic source of differentiation and advocacy.
The enduring job of marketing remains human: create meaningful value for people, even as the tools used to accomplish that job become increasingly intelligent.
Key Techniques
5D Framework: Monitor five drivers of market change: technology, political-legal forces, economy, socio-culture, and market dynamics.
4C Framework: Translate the landscape into four areas of strategic analysis: change, competitor, customer, and company.
Three Gateways: Evaluate marketing through social, personal, and experiential pathways into the customer’s mind.
Cognitive Mapping: Observe how customers process attention, social influence, rewards, trade-offs, and decisions to uncover deeper behavioral insight.
Cognitive Value Design: Examine how benefits, sacrifices, context, and biases shape perceived value rather than assuming customers calculate value rationally.
Journey Mapping: Examine the customer journey across discovery, hospitality, transaction, onboarding, support, and engagement, then identify the moments that deserve disproportionate attention.
WOW Moments: “Break the script” by delivering something unexpectedly positive during important or exceptional customer moments.
Multisensory Experience Design: Use sensory immersion, sensory harmony, and emotional memory to make experiences more distinctive and memorable.
Share-Worthy Experience Design: Build personal resonance, social recognition, and easy sharing into experiences to encourage advocacy and word of mouth.
Author’s Qualifications
Philip Kotler is Professor Emeritus of Marketing at Northwestern University’s Kellogg School of Management, where he held the S. C. Johnson & Son Professorship of International Marketing. He holds an MA in economics from the University of Chicago and a PhD in economics from MIT. His books have been translated into more than 25 languages, and he has spent decades shaping modern marketing education and practice.
Hermawan Kartajaya is founder and chairman of MCorp and a longtime Kotler collaborator. He has been recognized by the Chartered Institute of Marketing as one of the “50 Gurus Who Have Shaped the Future of Marketing” and founded organizations including the Asia Marketing Federation and World Marketing Forum.
Iwan Setiawan is Group COO of MCorp, a marketing consultant with more than 20 years of experience and work with more than 100 clients. He teaches marketing in the Executive MBA program at Bandung Institute of Technology and holds an MBA from Northwestern University’s Kellogg School of Management.
Comparison to Similar Books
Marketing 7.0 is best understood as the next step in Kotler, Kartajaya, and Setiawan’s Marketing X.0 series. Marketing 4.0 examined the connected customer and digital transition, Marketing 5.0 emphasized using advanced technology for human needs and personalization, and Marketing 6.0 explored immersive physical-digital experiences. Marketing 7.0 moves one level deeper by asking how all of those technologies interact with cognition and decision-making.
Compared with tactical AI-marketing books, it is much less about specific tools, prompts, or workflows. Compared with behavioral-science books such as Daniel Kahneman’s Thinking, Fast and Slow or Richard Thaler and Cass Sunstein’s Nudge, it is less academically deep but more explicitly organized around practical marketing functions such as positioning, storytelling, value, selling, and customer experience. Its distinctive contribution is connecting cognitive principles with the authors’ established strategic marketing frameworks.
Target Audience
Marketing executives and CMOs determining how AI should fit into marketing strategy
Brand strategists concerned about differentiation and authenticity in an increasingly automated market
Digital marketers who want to look beyond short-term performance metrics and optimization
Business owners and executives trying to understand changing customer behavior in the AI era
Customer experience, product, and sales leaders interested in how cognition influences value, trust, and advocacy
Marketing consultants, educators, and students who want an updated strategic framework for modern marketing
Experienced marketers looking for a human-centered counterbalance to the current emphasis on AI tools and automation
Critical Response to the Book
Early response to Marketing 7.0 has generally praised its timely human-centered perspective on AI. Reviewers have highlighted the concept of mind-centric marketing, its warning that AI-driven optimization can commoditize marketing output, and its insistence that technology should strengthen rather than replace human judgment.
The most consistent criticism is that the book is stronger as a strategic framework than as an execution manual. Some readers view its themes as an incremental continuation of the earlier Marketing X.0 books rather than a radical new theory, while others wanted more concrete AI tactics and implementation guidance. That criticism is fair, but it also reflects the book’s intended role: it is primarily a book about how marketers should think in an AI-saturated environment, not a handbook for operating particular tools or channels.
One Sentence Takeaway
To sum up: As AI makes marketing faster, cheaper, and easier to automate, sustainable differentiation will increasingly depend on understanding the human mind better than competitors understand the technology.
Publishing a new article doesn’t mean your existing visitors will find it.
Hitting Publish is really the beginning of the work. Your new article may get indexed by search engines, shared on social media, or included in an email, but you still need ways to put it in front of people who are already visiting your site.
That’s where Latest Category Article promotions in SiteSqueeze come in.
Instead of manually adding links to older posts every time you publish something new, you create the promotion once. SiteSqueeze uses your existing content to automatically surface the newest eligible article in the same category.
As you publish again, the article being promoted can change automatically.
Set It Up
From your WordPress dashboard, go to SiteSqueeze → Promotions and select Create → Latest Category Article.
Setup is minimal: name the promotion so you can find it later, toggle it on or off, and set the button text (defaults to “Read More”). Unlike a Standard Promotion, there’s no destination URL, headline, or image to configure. SiteSqueeze handles those dynamically.
How SiteSqueeze Picks the Article
This is the core difference from a Standard Promotion. You don’t choose the destination. SiteSqueeze finds the newest eligible published article in the applicable category the visitor is reading, then pulls its title, featured image, and URL.
Two rules govern eligibility:
No self-promotion. The article a visitor is currently reading is excluded.
Featured image required. If no other eligible article in the category has one, the promotion won’t display.
Say your newest Analytics article today is Understanding Attribution Models. That’s what gets promoted to readers of older Analytics posts. Publish 5 GA4 Reports Every Marketer Should Use next week, and it can take over as the promoted article automatically. No edits, no new URL, no swapped image.
Placements Control Where It Appears
Enabling the promotion isn’t enough on its own. It also needs a Placement.
The distinction: Placement decides where a promotion can appear; SiteSqueeze decides what gets promoted. A Placement tied to your Analytics category opens the door for the Latest Category Article promotion to run on eligible Analytics content, and SiteSqueeze fills in the newest eligible article.
Placements can use categories or tags to set eligibility, but article selection always follows category. If you’re using the All Others Placement, SiteSqueeze pulls from the first eligible category alphabetically.
One constraint: a Latest Category Article promotion must be the only promotion assigned to its Placement.
Start/End Date — schedule when it’s eligible to display
Impression Limit — stop it after a set number of views
Click Limit — stop it after a set number of clicks
None of these are required. If you want Latest Category Article running indefinitely as a standing way to circulate newer content, leave them off. If you set more than one limit, the promotion stops at whichever condition hits first.
Track What Actually Gets Promoted
Because the destination changes, a promotion showing 100 impressions doesn’t tell you the whole story; those impressions could be spread across several different articles. SiteSqueeze’s Latest Category Article Performance report breaks down which categories generated opportunities, which articles got promoted, and how visitors responded. I cover that in detail in SiteSqueeze Latest Category Article: Understand What Content Gets Promoted.
The Bottom Line
Every article you publish is another entry point to your website. Latest Category Article promotions make sure older entry points keep leading somewhere current, without you having to maintain the link. Set it up once, connect it to the right Placement, and it stays current as your content library grows.
SiteSqueeze is coming soon.
I’m building a WordPress plugin designed to help marketers promote their own content, measure what works, and run practical A/B tests without pretending every website has enterprise-level traffic.
AI search visibility is quickly becoming another thing marketers want to reduce to a score.
How often was my brand mentioned? How many citations did I earn? What’s my AI share of voice?
Those metrics matter. But a visibility score can tell you that you showed up without telling you how you got there, what AI said about you, or whether any of it influenced a decision.
Retrieval was the piece I hadn’t been thinking about clearly enough. Retrieval happens upstream of citation, but being retrieved doesn’t mean you’ll ultimately be cited. And even after your brand makes it into a response, one question remains: did that visibility actually matter?
That led me to five questions I think give a more complete picture of AI visibility, from technical opportunity to business impact:
Were you retrieved?
Were you mentioned?
Were you cited?
How were you represented?
Did that representation influence a decision?
They aren’t perfectly sequential stages, and I’ll get into why that matters. But together, they expose real gaps in how we currently measure AI search visibility.
1. Were You Retrieved?
Before you can worry about being cited, there’s a more fundamental question: did the AI system retrieve your content at all?
85% of the pages ChatGPT retrieved never became citations.
Retrieval creates an opportunity to be considered. It doesn’t guarantee your content survives the selection process.
This is also where traditional SEO and AI visibility overlap more than the “SEO is dead” narrative suggests. Many of the fundamentals marketers already work on, including technical accessibility, clear content structure, internal linking, semantic clarity, and authority, can still contribute to whether information is discoverable and usable.
But retrieval introduces a measurement problem: marketers usually can’t see it. A citation is visible. A brand mention is visible. Retrieval without either often isn’t — which makes it both foundational and one of the hardest parts of AI visibility to track.
2. Were You Mentioned?
Next comes the metric most AI visibility platforms already measure well: did your brand appear in the answer?
Mentions matter because they tell you whether you’re in the conversation at all. If someone asks an AI assistant for the best software in your category and your company keeps showing up alongside your competitors, that’s useful signal.
But a mention doesn’t tell you where the information came from. An AI system can mention your brand without citing your website — pulling instead from a third-party review, a community discussion, or information it learned during training.
That’s the gap between mention and citation. A mention answers did I show up? It doesn’t answer did my content help me show up?
3. Were You Cited?
Citations give you something mentions don’t: attribution. If an AI-generated answer links to your page as a source, you have stronger evidence your content actually shaped the response.
That’s why citations have become such a prominent AI search metric — they’re visible, countable, and can drive referral traffic.
But citations have limits too. If only 15% of retrieved pages became citations in the AirOps dataset, then citation tracking alone misses most of what happens earlier in the process. And being cited doesn’t mean you were prominently mentioned or recommended — your page might support one fact in an answer while a competitor gets the actual recommendation.
Citation matters. It’s just not the finish line.
4. How Were You Represented?
This is the question I think marketers underrate most: showing up is not automatically a good outcome.
An AI system can mention your company and still:
Describe your product incorrectly
Use outdated pricing or features
Misunderstand who your product is for
Compare you with the wrong competitors
Leave out an important differentiator
Repeat inaccurate information from a third-party source
If that happens, your visibility metric can look great while your actual brand representation is poor. The better question isn’t “did AI say something favorable about us?” It’s: was the representation accurate, current, complete, and appropriate to the question asked?
This pulls AI visibility into brand management territory, and it extends beyond your own website. AI systems pull from review sites, publishers, forums, comparison pages, and databases — your site can say one thing while the rest of the information ecosystem says another.
5. Did That Representation Influence a Decision?
Eventually marketing has to answer the harder question: did any of this cause someone to do something — search for your brand, visit your site, add you to a consideration set, request more information, buy?
This is where AI visibility becomes a business measurement problem, not just an SEO or GEO one. And attribution gets messy fast: someone discovers you through ChatGPT, searches Google later, reads reviews on Reddit, visits your site directly, and buys three days later. Analytics records a branded or direct conversion. The AI interaction that started it all disappears from the trail.
That’s why I’m skeptical of judging AI search’s business value purely by referral traffic. Referral traffic matters, but influence can happen without a click.
Visibility isn’t the goal. Influence is.
This Looks Like a Funnel, but It Isn’t
It’s tempting to line these up as a funnel: retrieved → mentioned → cited → represented → influenced. It makes a clean visual, but it’s technically imperfect. You can be:
Retrieved and never cited
Mentioned without being cited
Cited without a prominent brand mention
Represented without your own website being cited
Influential without producing a measurable site visit
There’s also a deeper wrinkle: content can be cited without being the primary information that actually shaped the answer.
So these aren’t five stages every AI answer moves through in order. They’re five questions marketers should keep asking as visibility moves from technical opportunity toward business impact.
What Can AI Search Software Actually Measure?
This is where the framework earns its keep. I’ve spent a lot of time testing AI search visibility software, and most platforms are far stronger in the middle of this model than at either end.
AI visibility question
What it tells us
Measurement today
Were you retrieved?
Whether your information entered the AI’s research and selection process
Difficult
Were you mentioned?
Whether your brand appeared in the answer
Relatively measurable
Were you cited?
Whether your source received attribution
Relatively measurable
How were you represented?
What the AI actually said about your brand
Measurable, but needs interpretation
Did you influence a decision?
Whether visibility contributed to a business outcome
Very difficult
Tracking prompts, mentions, citations, sentiment, and share of voice tells you a lot about what shows up in AI-generated answers. It doesn’t tell you why you got selected, or whether being selected actually mattered. Those may be the two most valuable questions in AI search visibility right now.
The tools are improving quickly, particularly around mentions, citations, sentiment, and share of voice. The harder problem is connecting those observable signals to what happened before the answer was generated and what happened after someone saw it.
AI Visibility Is Bigger Than Citation Tracking
I don’t need another argument over whether SEO, GEO, AEO, or some other acronym wins. What interests me is the actual visibility problem: can AI systems find and use our information, do we show up when relevant questions are asked, are we earning attribution, are we described correctly, and does any of that move a decision?
That’s also the gap between measuring AI visibility and improving it — two different jobs. A dashboard can tell me I wasn’t cited. The more valuable platform tells me why, and what to do about it. It’s why I treat AI search as one piece of a broader AI Visibility Framework rather than a standalone replacement for SEO, and why I keep coming back to these five questions when evaluating AI search software:
Knowing your brand showed up in an AI answer is useful. Understanding how it got there, what was said, and whether it mattered is what actually matters.
A/B testing in SiteSqueeze puts two promotions head to head so you can see which one actually earns more clicks. Pick two Standard Promotions, set a threshold for what counts as a real improvement, assign a Placement, and start the test. SiteSqueeze splits your traffic roughly 50/50 and tracks the results.
Control A: your current promotion, headline and image as-is
Variation B: an alternative version with something you want to test
If you want to understand why one version performs better, change one major variable at a time. Changing the headline, image, and call to action together can tell you which overall promotion performs better, but not which change made the difference. If your goal is simply to compare two different creative approaches, changing multiple elements can make sense.
Set Up the Test
A/B testing is a SiteSqueeze Pro feature.
From Promotions, select Run an A/B Test, then Create A/B Test.
Name the test something you’ll recognize later. “SiteSqueeze Headline Test – September” beats “Test 1” every time.
Then choose:
Control (A): the promotion you’re testing against
Variation (B): the challenger
These have to be two different promotions.
Set Your Minimum Meaningful Improvement
This threshold decides how much better one promotion needs to perform before SiteSqueeze considers the difference meaningful:
Threshold
Meaning
10%
Small improvement
20%
Recommended (default)
30%
Strong improvement
50%
Major improvement
A 2% lift might be real, but if it’s not big enough to change what you do next, it’s not worth treating as a winner. Set this threshold before the test starts, not after you’ve seen which version is ahead.
Choose how the test ends, if you want a hard stop:
Total impressions
Total clicks
End date
Leave these blank for no limit.
A stopping limit ends collection. It doesn’t force a winner. If a test hits 5,000 impressions without strong enough evidence, SiteSqueeze marks it Inconclusive rather than declare a winner it can’t back up. That’s by design.
Save, Then Assign a Placement
Click Save A/B Test. The test now exists, but it isn’t live yet.
SiteSqueeze routes A/B tests through Placements, the same system that controls every other promotion type, so all your targeting rules live in one place.
Go to Placements, create or edit the one you want, and select your A/B test. Target it by category, tag, or All Other Content.
One rule: while the test runs, Control A and Variation B can’t have their own separate Placements. The test needs full control of delivery.
SiteSqueeze now handles delivery: visitors are split roughly 50/50 between Control A and Variation B, and returning visitors see the same variation each time browser storage allows it.
You can Pause Test anytime and resume later without losing collected data. Stopping the test entirely preserves whatever statistics it already gathered.
What to Expect
Let it run. SiteSqueeze needs at least seven full active days of data before it will recommend a winner, and even then, seven days alone isn’t a guarantee the evidence is strong enough.
Setting up a test is the easy part. Deciding what’s worth testing is what actually matters.
Skip “which one wins” and ask something you want an answer to:
Does a benefit-focused headline beat a feature-focused one?
Does a product shot outperform a conceptual image?
Does “Download the Guide” out-click “Learn More”?
Flint McGlaughlin of MECLABS has a line I’ve always liked: “The goal of a test is to get a learning, not a lift.”
That’s the mindset to bring to SiteSqueeze. A winning variation is useful. Learning why it won gives you something you can carry into the next promotion, the next test, and the marketing decisions that follow.
SiteSqueeze is coming soon.
I’m building a WordPress plugin designed to help marketers promote their own content, measure what works, and run practical A/B tests without pretending every website has enterprise-level traffic.
