Category: Distribution

  • X Marketing Experiment: Does a Premium plan or follower count guarantee reach?

    X Marketing Experiment: Does a Premium plan or follower count guarantee reach?

    One of the most controversial changes Elon Musk made after his acquisition of Twitter was the monetization of its verification program. Back in the day, Twitter provided a blue check to celebrities, journalists, and those whothat it deemed at risk of impersonation. Musk argued that the blue check program produced a “privileged class” of users, and decided to open up verification to anyone willing to pay for it. This also would help the company develop revenue streams beyond advertising. 

    The initial pricing for verified plans was $8/month. Eventually, X developed a three-tier pricing structure, plus special pricing plans for business profiles. Additionally, verified users in the highest pricing tiers are able to monetize their tweets’ engagement, with payouts spanning from a few dollars to thousands. 

    Although originally Musk and the new Twitter leadership hoped that the new verified users would be less likely to incur in harmful behavior, since the program launched:

    • Verified users have been involved in memecoin scams, impersonation and other harmful behavior.
    • Thousands of engagement-baiting and automated profiles have bloomed, decreasing X’s average content quality. 

    However, some interesting content still lives on Twitter – you just have to find it. But as a brand – how can you make it easier for users to find you? Especially at this point in internet history, optimizing your content for discoverability is a favor you do to your audience, helping them find entertaining/informative/original content in a sea of AI-generated sameness.

    Three X profiles, three different stories and three different niches

    In Q4 2025, we monitored and experimented with three X accounts we managed:

    • Account A: A verified account started in 2013, with the highest-tier X plan, c.20k followers (including many renowned accounts that were verified before paid verification was rolled out). 
    • Account B: A new, public account with no blue checkmark and 28 followers. 
    • Account C: A private account with no blue checkmark, started in 2018, with 456 followers.

    Followed-follower ratio (the amount of users someone follows per follower they have) has historically been an indicator of content quality, as well as a signal of authenticity. It’s very common for bots to have a very high volume of followed accounts and little to no followers of their own.

    When it comes to followed-follower (f-F) ratio:

    • Account A has a f-F ratio of 0.04-1
    • Account B has a f-F ratio of 1.5-1
    • Account C has a f-F ratio of 0.5-1

    When it comes to activity levels:

    • Account A published an average of 1 tweet every other day.
    • Account B published with a frequency of under 1 tweet per week.
    • Account C published approximately 10 times a day.

    When it comes to topic coverage:

    • Account A mostly replies to high-engagement news articles with short insights and memes. Account A focuses on finances, world events and politics – three of X’s most active niches.
    • Account B mostly retweets or replies to lighter content, focused on relationships, personal anecdotes and dating. Original content is scarce.
    • Account C is a mix between the two.

    The three accounts published content in Spanish. None of the accounts feature the user’s legal name or other identifying information.

    Our goal was to test three hypotheses:

    • Does account age affect visualizations?
    • Does paying for the highest premium plan affect visualizations (as promised by X itself)?
    • What type of sentiment is currently producing the most engagement?

    Our findings

    X marketing experiment visualization results
    • As expected, Account C’s visualizations have been below its total number of followers. Considering that access to the account’s content is reserved for its followers, this is reasonable. Retweets were 0 across this account’s sample.
    • Account B saw the highest volume of visualizations.
    • In spite of the promised visibility boost and its medium-size follower base, Account A’s tweets were only shown to a small number of users. 
    X marketing experiment likes results
    • Account B was a winner, with like counts in the third digits (vs. Account A & Account C).
    • General engagement was consistent with the volume of visualizations. More visualizations = more volume.
    • If we track engagement-follower ratio, Account C gets the lead, with 10 engagements per 100 followers.

    So, what does this mean?

    • A historied account with a large follower count, a healthy followed-follower ratio and a moderate publishing volume doesn’t have guaranteed distribution. What worries us isn’t engagement rates but low distribution. Without a high-enough visualization volume, high engagement rates are impossible.
    • Publishing frequency doesn’t seem to be a factor.
    • Non-specialized, relationships & pop culture-related content has a higher potential for distribution and engagement than political content. This may be an attempt by X to attract new users and distance itself from its reputation as a politically problematic platform.

    Why Sports Brands Outperform Everyone on Twitter

    Despite overall declining engagement on X, the sports niche remains a notable exception where marketers can still find relatively high user engagement. In the past year, sports teams have kept one of the highest interaction rates in the platform. In 2024 and 2025, sports content achieved an average engagement rate around 0.07%, more than double the platform average of about 0.029%. This contrasts with media brands, which experience much lower engagement rates (near 0.009%). Sports teams also post frequently—around 44 tweets per week—helping maintain strong audience interaction compared to other sectors. 

    Let’s explore some potential reasons behind this anomaly.

    • Frequent, real-time posting: Sports teams post very actively—around 44 tweets per week—keeping fans constantly updated with live scores, highlights, and news, which sustains ongoing conversations and engagement. Twitter was great for tracking real-time events and X still is. However, it’s worth considering that, as our experiment shows, content volume alone doesn’t bring results.
    • Dedicated content portals and aggregation: Leagues like the NBA and NFL have launched dedicated portals on X that aggregate content from teams, players, and media, creating a centralized hub for fans to follow and interact with real-time updates and exclusive video highlights.
    • Community and emotional investment: Sports fans are highly engaged and emotionally invested in their teams, making them more likely to interact with content through likes, retweets, and replies. The platform’s algorithm favors tweets with high engagement, amplifying visibility for sports-related posts.
    • Preference for fresh, time-sensitive information: X users favor informative, relevant, and trending content. Sports content naturally fits this demand by providing timely, newsworthy updates and participation in trending conversations, which drives higher engagement.
    • Multimedia content boost: Tweets with videos and images, common in sports posts, attract significantly more engagement. Video views and completion rates help keep fans engaged longer, enhancing interaction metrics.
    • Year-round fan engagement strategies: Sports organizations are increasingly offering immersive, interactive experiences beyond game days, such as behind-the-scenes content, live draft coverage, and exclusive athlete streams, keeping fans connected throughout the year.

    Does X Premium increase reach?

    Across large datasets, it seems to be the case. Buffer analyzed 18.8 million posts from 71,000 accounts between August 2024 and August 2025 and found median Premium reach near 600 impressions per post, Premium+ above 1,550, and regular accounts under 100.

    But Buffer’s study is observational – and it’s hard to determine what came first, if it was the chicken or the egg. The study compares groups rather than randomizing them, so people who pay may also post more or better. Are good posters simply more likely to buy a premium X subscription? The Buffer study also measures medians, which hide individual outliers like our Account A. Our result fits that picture only if something other than the subscription dominated Account A’s reach. The next sections cover what could.

    A look inside the X algorithm: How does X rank posts in 2026?

    X published the code behind its For You feed on January 20, 2026, then added the scoring parameters and visibility-filtering systems on August 13 and 14. The ranking scorer multiplies the model’s predicted probability of each user action by a weight, then sums the results.

    According to one reading of the public parameter file, the defaults in param.rs put a like at 0.5, a repost at 1.0, a reply at 5.0, and a share by copied link at 20.0. These weights scale predicted probabilities, not raw engagement counts, and X added comments to its own repository saying so. Read them as “what the model values,” not “how many likes a reply is worth.”

    Which mechanisms could explain our results?

    These are hypotheses drawn from the repository, not findings. Our test ran in Q4 2025 and the code is from 2026, so the code describes the current system, not necessarily the one we measured.

    Mechanism in X’s codeAccount it may explainHow to test it
    Replies and reposts from accounts a viewer doesn’t follow are removed before scoring in For YouA and B (both mostly replies and retweets)Compare views on replies vs. original posts from the same account
    A new-author boost lifts posts from authors below an impressions threshold, per the ranking scorerBTrack B’s views as its impression history grows
    Visibility filtering takes account of whether an account is protectedCNot testable without a public twin account
    Each additional post from the same author is multiplied by a decaying factor, and posts older than 48 hours are droppedC (about 10 posts a day)Vary posting cadence, hold topic fixed

    What should operators measure first?

    Measure eligible reach first, then response, then business conversion, then what you learned. A post that is never shown cannot generate a meaningful engagement test, and an engagement rate on 40 views proves little about the channel.

    LayerQuestionMetric
    DistributionWas it shown to enough relevant people?Views per post, segmented by format and topic
    ResponseDid exposed users react?Replies, link-copy shares, reposts, profile visits per 1,000 views
    ConversionDid attention produce an action?Tagged signups, demos, subscriptions
    LearningWhat does the next post test?A written hypothesis

    How should you test X distribution yourself?

    Run one public account for four weeks and change one variable at a time.

    QuestionDesignDecision rule
    Does topic matter?Alternate two topics, same format and windowKeep a topic only if it lifts qualified views across several posts
    Does format matter?Same idea as short text and as mediaKeep the one with more qualified actions per view
    Does verification matter?Compare matched posts before and after a plan changeProvisional until it repeats
    Do replies beat originals?Same thesis as an original and as a replyChoose by profile visits or tagged actions

    Pre-register the question, archive the raw export before charting, and publish the dataset.

    What would change our conclusion?

    If a controlled test shows verified accounts reliably out-reaching matched unverified accounts, we will say Premium buys distribution and report the size of the effect. Until then, treat verification as a credibility and feature choice, and treat reach as something you measure per account.

    Is Twitter marketing still worth it in 2026?

    We leave that for you to decide! But, in spite of worrying metrics, high bot activity and a reputation for unhinged and radical political content, Twitter still has plenty to offer to both brands and users. Maybe, we’re on the eve of a new era for the platform, powered by real-time pop culture coverage and fan culture. Will Twitter be able to recover its former glory? Only time will tell.

    FAQ

    Does this prove X Premium doesn’t increase reach?

    No. It shows one verified account in a small, uncontrolled sample did not out-reach two other accounts. Buffer’s 18.8-million-post study finds a large average Premium advantage, so a controlled test is needed to separate the subscription from posting habits and content.

    Do link posts get penalized on X?

    The evidence conflicts. Buffer found link posts from non-Premium accounts at roughly 0% engagement from March 2025, but Elon Musk said in July 2026 that X had not penalized links for over a year, and PPC Land’s review of the released code found no URL-targeting rule. Posting the substance natively and putting the link in a reply is the lower-risk approach.

    Is X’s algorithm really open source?

    Substantially. The repository is public under an Apache 2.0 license, but some pieces are deliberately withheld, including the Grok prompts used for content classification, and live parameter values can differ from the repository defaults.

    Should brands ignore follower count?

    No. Followers are a reachable audience record, but they don’t forecast how far any single post travels.

    By Aaron Marco Arias

  • HubSpot is fine. HubSpot SEO is not.

    HubSpot is fine. HubSpot SEO is not.

    It’s October 2026. Ahrefs estimates HubSpot’s organic traffic at ≈3.6 million visits, down roughly −689K month on month. The platform also places the site far below its 2023 and early-2024 peak of almost 11 million visits. The number is a third-party estimate, not HubSpot reporting, but similar platforms point in the same direction: HubSpot’s organic traffic isn’t what it used to be.

    Some would say that this is completely natural, since “SEO is dead.” But is that true?

    HubSpot remains a powerful company with a powerful brand, they’ll be fine. But the model it taught B2B software companies to copy – publish answers to every informational question and let traffic compound into demand – depended on conditions that search no longer supplies for free.

    What does the data tell us about HubSpot’s current SEO standing?

    HubSpot's performance and traffic trends (Source: Ahrefs)

    Back at its traffic peak, in February 2024, HubSpot ranked for over 8 million keywords. Most of these keywords were in very low positions, with over 2 million keywords in positions 21-50 and almost 3.5 million beyond position 51.

    Fast-forward to the present day and HubSpot is only ranking for 278,294 keywords, with only two organic positions beyond spot 51.

    HubSpot's performance and traffic trends (Source: Ahrefs)

    HubSpot is the brand that defined and launched the concept of inbound marketing. They’re a category-shaping leader within the highly concentrated CRM market. But, even with those factors in mind, 8 million keywords is a bit too much.

    Are there 8 million inbound-marketing and CRM related keywords that HubSpot fought to dominate, or was there something else going on?

    HubSpot’s inbound bargain

    If we filter out branded keywords, we’ll find that some of HubSpot’s organic traffic comes from keywords that have nothing to do with their product – "excel online" brings over 206,000 visitors per month, while "how to use instagram" brings over 209,000.

    HubSpot's non-branded top keywords snapshot (Source: Ahrefs)

    The SEO playbook that HubSpot relied on aimed towards volume rather than product-content fit. Generic, tech-related keywords were perceived as free real estate to cover for awareness purposes. The query "car insurance quotes" doesn’t suggest intent to buy a CRM subscription.

    HubSpot's organic pages trend (Source: Ahrefs)

    The peak in HubSpot’s organic traffic coincides with their peak in organic pages. Since then, the website has experienced a reduction in content volume. Could it be that HubSpot is cleaning up their enormous content library and focusing on its strongest content? Was HubSpot’s problem a large volume of off-topic content?

    The content-led growth playbook requires that we make certain assumptions – do they still stand in 2026?

    Let’s run with that hypothesis for a while: Content-led growth worked because a company could turn broad informational demand into an owned audience. That bargain relied on four conditions:

    ConditionWhat it suppliedWhat is changing
    Click surplusA ranking produced a visitAnswer layers, richer results and AI interfaces increasingly absorb “what is” & “how to” questions
    Authority transferA strong domain could publish broadlyHigh authority now coexists with less traffic – HubSpot has a DR of 93/100, and it’s still bleeding traffic
    Scope toleranceSearch rewarded volume outside the core productBroad, lightly related libraries face higher quality and relevance pressure
    Attribution forgivenessSessions could stand in for commercial impactBrand, qualified demand and pipeline now tell a different story from raw traffic

    We can see the same trends across certain pockets of the media ecosystem, with publications increasingly relying on video + social traffic, while losing big on the SERPs. Of course, there are exceptions – both the New York Times and the Wall Street Journal are enjoying growing organic traffic (albeit with fewer keywords).

    So, back to HubSpot – this hypothesis outlines a very clear story: Concentrated content, and fewer pieces with true market fit outperform spread-out, generalist content. Great!

    So, remember that “car insurance quote” example we brought up when discussing HubSpot’s questionable SEO practices? We’ve got something awkward to tell you…

    What’s the impact of the HubSpot SEO playbook?

    We tracked down which page HubSpot’s using to rank for "car insurance quotes". It has nothing to do with car insurance. It’s a very technical and valuable guide on how to create quotes on HubSpot’s legacy quote creator. We checked its backlink history and noticed that it’s a permanent redirect from a different URL. Ha! Maybe the old content was full of off-topic advice about car insurance quotes.

    Mmm, no. The old guide was very similar to its successor and it had nothing to do with car insurance. The terms “car” and “car insurance” are fully absent from the content and its metadata. So, what’s going on?!

    HubSpot’s knowledge base article on configuring legacy quotes (/quotes/set-up-legacy-quotes) ranking for "car insurance quotes" provides a fascinating case study in how modern search engines process semantic signals across large domains. While the documentation page itself contains standard B2B software setup instructions for e-signatures and pricing settings, the root cause of this ranking stems from user-generated content on community.hubspot.com. There, a forum thread specifically asked how to automate quote calculations for a car insurance company, explicitly tying the concepts of “car insurance,” “quotes,” and “HubSpot” together on the domain.

    Modern search engine algorithms rely heavily on dense retrieval and vector embeddings to group related concepts, mapping connections across an entire site rather than evaluating pages in total isolation. By combining the entity signals from the user-generated community forum with the structural authority of the official documentation page, Google’s semantic models created a conceptual bridge between “car insurance” and HubSpot’s native quoting tool. Paired with HubSpot’s immense domain authority, the algorithm over-generalized these semantic signals, leading Google to mistakenly surface a technical software setup guide for a high-volume consumer auto insurance query.

    What does this mean?

    This specific anomaly is an accidental microcosm of the exact mechanism that triggered HubSpot’s broader SEO collapse.

    HubSpot’s long-standing organic playbook relied on building massive domain authority and casting an extremely wide net, publishing thousands of top-of-funnel articles on generic subjects to capture maximum web traffic and funnel those visitors toward their CRM products.

    This “car insurance” example directly illustrates how that “wide rather than deep” strategy backfired. This is happenning because Google doesn’t really understand the HubSpot website’s scope.

    The useful distinction is between reach and demand

    For years, ranking for tangentially related terms was considered a free win. But as search engines transitioned from keyword matching to strict topical boundaries, those fringe rankings went from being free brand awareness to actively poisoning the domain’s topical focus.

    The old playbook collapsed two different jobs into one metric. A post could reach a huge audience and still create little demand for the product. That ambiguity was tolerable while search delivered cheap clicks at scale. But as organic traffic dries up and budgets tighten, SEOs are starting to say the quiet part out loud: This type of strategy was always an awareness play with very weak ties to conversion.

    HubSpot will survive. Beyond their SERP presence, they’re backed by years of narrative capital. HubSpot’s best keywords are branded, and a user searching for “HubSpot pricing” or “HubSpot Academy” has already crossed a threshold that a generic “what is CRM?” visitor may never cross. The company can build from that advantage. Smaller companies that copied the publishing model usually cannot.

    New times demand new models.

    AI citation is not a rescue plan, welcome to the zero-click era

    We’re in the age of “zero-click searches” – basically, user journeys that begin and end on the Google SERPs, without users ever moving forward to a third-party page. With AI overviews, Google covers most informational queries, capturing traffic that, before this feature, would have gone to publishers.

    This is one of the main reasons why so many websites are losing traffic, HubSpot included. Not only was HubSpot’s content unrelated to its product, it was also far too generic and TOFU (top-of-funnel). Let’s take a look at a concrete example:

    Recently, HubSpot lost their positioning for the keyword "what is a target audience".

    Back in the day, an article titled “What is a Target Audience?” was amazing for attracting curious users who would be taken on a learning journey by HubSpot. The amazing value provided by the article would create a positive association in the user’s mind: HubSpot is where quality marketing info’s at. And this user would be a click away from discovering the product, signing up for a free trial, converting… they may not do it today, but eventually, when they need a CRM, they may remember that HubSpot taught them what a target audience was.

    That’s what the same experience looks like today:

    Google AI Overview screenshot, "what is a target audience?"

    The user searches “what is a target audience” and receives an AI Overview, at the top of the SERPs, providing them with all the information they need to know.

    Are you in a HubSpot SEO-like situation? Here’s what to do

    1. Audit and clean up your content library

    If your company is in HubSpot’s SEO situation – that is, with an inflated and exceedingly diverse content library, audit your website and delete all off-topic content. Organize the remaining content into topical clusters.

    Concentrate on bringing traffic to a couple of pieces with high content-product fit. In fact, you may find that your best pieces may be long due for an update.

    That’s a great start, but it’s far from enough. What should you do about the AI Overviews’ intercession between your users and your brand?

    2. Move TOFU out of SEO

    If a user’s not asking Google, they’re probably asking ChatGPT or Claude. Of course, you can always get ranked as a reference within AI search, and AI can bring you real traffic and customers.

    But, to keep the ROI of every piece as high as possible, move your long-form written content down the funnel. Your SEO TOFU motion is being sabotaged by AI, so move TOFU to social and use our long-form content for MOFU and BOFU. Make sure you’re covering long-string keywords that suggest high purchasing intent: Your product comparisons and product-centric tutorials. Want to keep some TOFU content on your website? Create brand-building, storytelling content and distribute it on social.

    3. Set up clear guardrails to prevent topical dillution from happening again

    How can you make sure that no new content suffers from low product-content fit?

    Before approving an idea, run it through this checklist:

    [ ] Product adjacency

    Can you name the feature, workflow or customer problem this page serves in one sentence? If the connection seems too convoluted or implausible, drop it.

    [ ] Unfair evidence

    Does the company have something a generic publisher cannot get: customer data, product access, benchmark results, support logs, practitioners willing to be quoted? If not, the page competes on effort alone, which is the part of the market answer engines now supply for free.

    [ ] Audience overlap

    Estimate what share of the query’s searchers could realistically become customers. A small number is acceptable for a brand play; it is not acceptable as the default assumption.

    [ ] Post-answer value

    If an AI summary satisfied the query, what would the visitor still want from the page? Calculators, templates, comparisons, pricing detail and screenshots survive summarization. Definitions and listicles do not.

    [ ] Path to demand

    State the next step after the visit – trial, demo, docs, comparison, newsletter – and where it appears on the page. No plausible next step means the page is a destination, not part of a funnel.

    [ ] Cluster membership

    Which existing topic cluster does it extend? A page that maps to no cluster and no commercial topic creates an island, and islands are what get pruned first.

    [ ] Metric and review date

    Name the metric that will judge the page in 90 days and the date someone will look at it. Sessions alone disqualify the page.

    [ ] Named owner

    Every page gets a person responsible for refreshing, consolidating or retiring it. Orphaned content is how the long tail fills with pages nobody defends.

    Frequently asked questions

    Is HubSpot’s traffic decline proof that SEO no longer works?

    No. The evidence supports a more precise conclusion: informational rankings do not guarantee clicks or commercial value. Search still matters where intent, product fit and page type align.

    Should companies stop publishing informational content?

    No. They should stop treating volume as a strategy. Publish when the company has distinctive evidence, expertise or utility to add, and measure the effect on demand rather than sessions alone.

    What should replace traffic as the main content metric?

    Use a mix of branded search, qualified pipeline influence, commercial-page engagement, conversion by entry point and retention of the audience you own. The right mix depends on the buying motion.

    By Aaron Marco Arias