The Algorithm Changed. That Does Not Mean Optimization Failed.

By Dr. Trudy Beerman, DSL — Published September 2, 2026

If you want someone else to introduce you to their audience, you need to give them a reason to make the introduction. That principle applies whether the gatekeeper is a television network, conference organizer, journalist, search engine, social platform, podcast host, or artificial intelligence system. The technology may be different, but the underlying question is remarkably similar: Why should this person, company, idea, or piece of content be selected for introduction to an audience that does not already know it?

I recently came across a LinkedIn post reacting to reports that ChatGPT had dramatically reduced how frequently it cited Reddit. The author's conclusion was essentially that this demonstrated why businesses should never hire an SEO, GEO, AIO, or whatever-the-next-acronym-is expert. When I saw the post, it had received 181 reactions, which interested me because the argument was clearly resonating with a meaningful number of people.

I agree with part of the skepticism. Anyone promising that they possess a permanent formula for controlling Google, ChatGPT, Bing, or another platform deserves scrutiny. Where I disagree is with the conclusion that changing algorithms somehow make optimization meaningless. I believe the opposite is true: the fact that selection criteria change is precisely why optimization requires ongoing observation, measurement, and adaptation.

Recommendation Is an Act of Selection

I run a television platform, so let me take this out of the world of algorithms for a moment. PSI TV has an audience, and when someone wants access to that audience through our programming, PSI TV determines what qualifies for distribution. We establish programming requirements, determine whether content fits the audience, decide what standards must be satisfied, and retain the right to change those requirements as the platform develops.

Someone who wants access to the PSI TV audience does not automatically have a right to that audience. They can decide that our requirements are not worth meeting and pursue another platform. They can build their own audience. They can decide television distribution is irrelevant to their goals. All of those are legitimate choices, but if they want PSI TV to make the introduction, then our current selection criteria matter.

Now apply the same principle to Google. Google has an audience. YouTube has an audience. LinkedIn has an audience. A conference organizer has an audience. A journalist has readers, a podcast host has listeners, and an AI assistant increasingly has users asking it to identify companies, experts, products, ideas, explanations, and sources.

Recommendation is an act of selection, and selection requires criteria. Once a platform takes responsibility for deciding what its audience will see, hear, read, or discover, some mechanism must determine what gets selected and what does not.

You Are Not Entitled to Someone Else's Audience

This is where I think much of the frustration with algorithms becomes misplaced. None of us is automatically entitled to Google's audience simply because we built a website. We are not entitled to YouTube recommendations because we uploaded a video, LinkedIn distribution because we published a post, a journalist's readership because we sent a press release, or a conference audience because we submitted a speaker application.

If we want one of those platforms to introduce us to people who do not already know us, a more productive question than "Why isn't the algorithm showing me?" is: What evidence have I provided that gives this platform sufficient reason to select me? That shifts the conversation from entitlement and luck to evidence and eligibility.

Google Gets to Decide What Google Shows

Google is quite open about the fact that its ranking systems evaluate many signals and that those systems evolve. Its Search Central guidance tells publishers to focus on useful, reliable, people-first content rather than creating content primarily to manipulate rankings. Google specifically asks publishers to consider whether their content demonstrates first-hand expertise and depth of knowledge and whether an intended audience would find the material useful even if they came directly to the website rather than finding it through search. Google Search Central: Creating Helpful, Reliable, People-First Content.

That last point is particularly relevant to why I continue writing articles like this one. I am not writing this because an SEO tool told me I needed another keyword-rich page. I encountered an argument, thought the conclusion deserved examination, and wanted to document my thinking because someone else wrestling with the same question may eventually find it useful. Whether Google sends that person to this page is Google's decision, but the intellectual asset exists regardless.

Microsoft provides an even more direct example of the relationship between traditional optimization and AI discovery. Its current Bing Webmaster Guidelines state that SEO fundamentals supporting discovery, indexing accuracy, and content clarity also support eligibility for AI-generated experiences, grounding results, and citations. Microsoft is not describing AI visibility as the death of optimization; it is explicitly connecting established search foundations to newer AI-driven forms of discovery. Bing Webmaster Guidelines.

If We Choose to Play, We Need to Understand the Current Rules

We do not have to optimize for Google. We do not have to publish on YouTube, participate on LinkedIn, pursue television exposure, pitch podcasts, seek media coverage, or care whether an AI system cites our work. A business is perfectly free to determine that a particular audience or distribution channel is not worth the effort required to reach it.

But if we decide that access to an audience matters, then understanding the platform's current selection environment is part of the work. We cannot reasonably demand that Google preserve yesterday's criteria because yesterday's criteria were convenient for us, any more than someone applying to appear on PSI TV can demand that I preserve an old programming requirement after the network's needs have changed.

We do not have to play the game with that platform. But if we choose to play, we should understand the current rules rather than complain that the rules can change. And we should expect them to change again.

Reddit Is Valuable, but Conversation Is Not Automatically Evidence

I also think the Reddit example deserves nuance because my disagreement with the conclusion does not mean I believe every source should carry equal evidentiary weight. Reddit can be enormously valuable for conversation, audience intelligence, experience sharing, questions, frustrations, recommendations, and discovering how people naturally talk about a subject. Those are legitimate forms of information, and they can reveal things a formal survey or corporate website may miss.

But conversation is not automatically factual authority. A Reddit contributor may be extraordinarily knowledgeable, but that same contributor may be mistaken. Their identity or qualifications may be unknown, a factual claim may not have been independently verified, and one person's experience may not establish a general truth. I therefore see nothing inherently unreasonable about a retrieval system distinguishing between a useful conversation source and a stronger primary or authoritative source when answering factual questions.

The more interesting question is not whether Reddit citations rise or fall during a particular period. It is what changing citation patterns teach us about the selection environment. That distinction matters because observing a change is different from claiming we know exactly why it happened.

Microsoft makes this limitation explicit in its own AI Performance documentation. Bing Webmaster Tools now allows publishers to observe which pages are cited in supported AI-generated experiences and which grounding queries are associated with those citations, but Microsoft cautions that changes in citation volume can result from multiple factors, including changes in user questions, content updates, and system or model updates. It specifically warns that these observational trends should not be attributed to one cause without evidence. Bing Webmaster Tools: AI Performance.

If Your GEO Strategy Was "Get Mentioned on Reddit," You Did Not Have a GEO Strategy

Suppose someone noticed that generative AI systems were frequently citing Reddit and decided the new visibility strategy should be getting brands mentioned there because "ChatGPT loves Reddit." If the citation pattern subsequently changed, that strategy might collapse almost overnight. But that would not prove GEO itself was fraudulent; it would demonstrate the danger of confusing a temporary tactic with a durable strategy.

If your GEO strategy was "get mentioned on Reddit because ChatGPT likes Reddit," you did not have a GEO strategy. You had a tactic. Tactics can be useful, but they exist within a particular environment and can lose value when that environment changes. Strategy has to account for that possibility.

Microsoft itself now uses the term Generative Engine Optimization in describing its AI Performance tooling. In February 2026, the company introduced the reporting capability as an early step toward GEO tooling that helps publishers understand how their content participates in AI-generated experiences. The report includes citations, cited URLs, grounding queries, and citation trends over time, which makes the idea that GEO is entirely imaginary increasingly difficult to defend. Microsoft Bing: Introducing AI Performance in Bing Webmaster Tools.

Optimization Does Not Force a Recommendation

There is another misconception worth correcting. Optimization does not mean control. A website can be well optimized and still not rank first. An outstanding speaker can submit a strong proposal and still not be selected for a conference. A credible expert can pitch a journalist and still not be quoted, and an excellent television guest can still be wrong for a particular program.

The same principle applies to AI systems. Strong content, credible evidence, clear authorship, appropriate structure, and recognized expertise may improve the conditions under which material can be discovered and considered, but they do not create an entitlement to citation. Optimization does not force the recommendation. It improves your eligibility for consideration.

This distinction is important because the selector retains the decision. Whether that selector is an editor, producer, conference committee, consumer, search ranking system, recommendation algorithm, or artificial intelligence system, the fundamental question remains: What evidence gives the selector sufficient confidence to introduce this person, brand, content, product, or idea to its audience?

This Is the Territory of Recommendation Science

This is one of the reasons I have become increasingly interested in what I call Recommendation Science. I use that term to describe the study of the conditions under which people, brands, ideas, products, and content are discovered, evaluated, selected, introduced, trusted, and recommended. The subject is larger than SEO because search engines are only one kind of selector, and it is larger than artificial intelligence because humans have been making recommendations for as long as humans have had choices.

Editors recommend. Journalists recommend. Television producers and podcast hosts recommend. Conference organizers recommend speakers, consumers recommend products, voters recommend candidates through their votes and advocacy, and communities help ideas and movements travel. Search engines and artificial intelligence systems introduce new technological mechanisms into an old human problem: deciding what or whom deserves to be put in front of someone else.

Microsoft's description of the emerging AI web reinforces why this deserves study. The company describes AI agents as increasingly acting as retrievers and being drawn toward structured, verifiable, and applicable content. Microsoft argues that visibility in this environment increasingly concerns how content contributes not only to rankings and clicks but also to answers, citations, reasoning, and outcomes. Bing Search Blog: Elevating the Role of Grounding on the AI Web.

REACHology® Is About Influential Reach

This developing field also helps explain why the name REACHology® continues to become more meaningful to me. Merriam-Webster defines an "ology" as "a branch of knowledge" or "science," which is remarkably close to what I intended when I began examining the conditions surrounding influential reach. Merriam-Webster: Ology.

REACHology® is not simply the study of reach as attention, impressions, views, followers, or virality. A person can attract enormous attention while having little substantive influence. Nor am I interested only in influence as persuasion that never travels beyond the people who already know the person or idea.

The territory I am interested in is the intersection: influential reach. It is the space where an idea travels and retains enough relevance, credibility, authority, or persuasive power to produce an effect. That effect might be a sale, an invitation, a vote, adoption of an idea, expansion of a movement, a changed behavior, or simply someone deciding, "You need to meet this person."

Reach asks whether the message traveled. Influence asks whether the message had the capacity to affect something. Influential reach asks what happens when both occur. That intersection is where movements expand, ideas spread, products sell, reputations grow, elections are won, experts become known beyond their existing circles, and recommendations introduce people to opportunities they could not access through their immediate networks alone.

A Science Does Not Pretend the Questions Are Finished

One of the things formal research taught me is that serious inquiry does not end by pretending every question has been answered. When I completed my own research, I was required to acknowledge the limitations of the study and identify areas that future researchers could investigate. The research itself revealed additional questions, potential relationships, and new rabbit holes that were outside the scope of the study I had completed.

That is part of what makes a field of study valuable. Observation generates questions, questions lead to investigation, investigation produces findings, and findings reveal patterns that can be tested, challenged, refined, replicated, or contradicted. A new finding is not necessarily the end of the inquiry; quite often it tells us what needs to be studied next.

I believe Recommendation Science should develop in the same spirit. We can observe what appears to increase discoverability, credibility, selection, and recommendation. We can measure what is measurable, document what changes, acknowledge limitations, test whether patterns continue, and revise conclusions when better evidence becomes available.

That is why an algorithm changing does not invalidate the study of optimization. Change creates more to study. If a recommendation system begins treating one type of evidence differently, the appropriate response is to observe the change, investigate it, measure what can be measured, identify what we do not yet know, and determine whether the findings can be replicated.

Understanding the Times Is Part of the Work

Scripture provides an interesting parallel in its description of the men of Issachar, who were recognized as men "who understood the times and knew what Israel should do" (1 Chronicles 12:32). Their distinction was not merely that they understood what had worked previously. Understanding the times meant correctly perceiving the environment they were actually in and knowing how to respond within it.

There is another biblical principle that is especially appropriate when thinking about platforms and opportunity. Paul wrote that "a great and effective door has opened to me, and there are many adversaries" (1 Corinthians 16:9). An open door represented opportunity, but opportunity did not mean the surrounding conditions were effortless, permanent, or under Paul's control.

The same principle of discernment applies here without turning algorithms into something more important than they are. If the audience changes, understand it. If the platform changes, understand it. If technology changes, study it. If the criteria for recommendation change, investigate what changed and determine whether reaching that audience still matters enough to adapt.

The Better Question Is Not "How Do I Beat the Algorithm?"

I think digital marketing has spent too much time teaching people to imagine themselves in a battle against algorithms. "How do I beat Google?" "How do I hack the algorithm?" "How do I make ChatGPT mention me?" Those questions encourage the very wizardry mentality that causes people to become cynical when a tactic stops working.

A better question is much more substantive: What legitimate evidence should exist so that a person or system evaluating me has good reason to conclude that I belong in the conversation? Sometimes the answer will involve SEO, GEO, or AEO. Sometimes it will involve media appearances, publishing, credentials, experience, reputation, original research, third-party validation, or a consistent digital footprint. Usually, meaningful authority is built through several of those signals working together rather than one isolated trick.

That is also why I continue writing. Not every thought I document needs to become a number-one Google result to have value. These articles create a record of what I am observing, questioning, learning, and developing as the environments surrounding influence, reach, authority, discovery, and recommendation continue to change.

Perhaps someone searching for an answer will eventually find one of these articles. Perhaps a future study will prove one of my observations incomplete. Perhaps several of these ideas will eventually connect in ways I cannot yet see. That possibility is not a reason to stop documenting the work; it is one of the reasons to keep doing it.

Recommendation is not magic. It is selection. If we want to be selected for introduction to new audiences, our job is to understand what gives the selector a credible reason to make that introduction. And when the selection environment changes, we do what serious students of any evolving field should do: observe, measure, learn, and adapt.