Sill Now Connects to Every Common CMS — and to Sites With No CMS at All
Read the postDateMar 27, 2026
CategoryProduct
Related reading
You can see your AI Share of Voice climbing in some places and falling in others. You can see which competitors are pulling ahead and which of your pages are earning the most citations. The harder question is what to do about it on Monday morning, and how to prove that the work you did actually made a difference. Recommendations is the new layer in Sill that turns your monitoring data into a clear, prioritized plan of work and routes every action you complete through the experimentation engine, so you know whether the change actually moved your visibility on the AI platforms that matter to you. Monitoring tells you what is happening. Recommendations tells you what to do about it. Experiments tell you whether it worked.
TL;DR
Sill's new Recommendations turns your monitoring data into a clear, prioritized plan of work. You get a queue of best-practice tactics that ship in days and strategic content or off-site campaigns that play out over weeks, scheduled into a four-week calendar that paces to how much work your team actually ships. When you mark a recommendation as done, Sill automatically measures whether the change moved your visibility on each AI platform, with statistical controls that separate real lifts from noise. Tactics that worked on your brand get prioritized in your next sprint; tactics that fell flat move down. Negative results surface a one-click recommendation to revert the change. Every recommendation comes with the published research that made it a priority and an expected lift you can defend to your team. Monitoring tells you what is happening, Recommendations tells you what to do about it, and experiments tell you whether it worked.
Pages updated within 90 days earn 67% more AI citations (SE Ranking, 2025), but most teams have no system that connects publishing decisions to the visibility outcomes that follow.
Most marketing teams running an AI visibility program live with the same uncomfortable gap. Their dashboard shows that visibility went up or down this week. Their CMS shows what was published. Nothing sits between the two. Decisions about what to write next get made from gut feel, the loudest competitor of the moment, or whichever advice article was in the team Slack on Friday. The work happens, the visibility moves, and nobody can confidently say whether one caused the other.
The cost of that open loop compounds quickly. AI platforms re-rank citation sources constantly; the same prompt that surfaces your brand today may surface a competitor tomorrow for reasons that have nothing to do with anything you shipped. Without a way to tie a specific change to a specific visibility outcome, teams end up either celebrating noise or abandoning work that was actually paying off. The fix is a layer that names the action you should take, schedules it into a sprint your team can actually run, and verifies whether it moved your visibility on each AI platform after you ship it. That layer is what Recommendations gives you.
Sill turns every signal already in your dashboard, from visibility gaps to perception shifts to factual errors AI platforms make about your brand, into specific actions you can run this week.
Your Recommendations queue is the answer to “what should I do next?” — a ranked list of actions tailored to your brand, your queries, and the AI platforms where you have the most to gain. Each one tells you exactly what to change, which page or venue it targets, why it was prioritized, and what kind of visibility lift you should expect from it. You can read the supporting evidence in one click, dismiss the items that do not fit your business, and complete the ones that do.
The queue is persistent, so the work you have already done does not disappear from view. You can see what is scheduled for this week, what is queued for next month, what your team has already shipped, and which recommendations have been measured against real visibility outcomes. When you tell Sill that a recommendation does not fit your brand, it remembers; when you tell Sill that a recommendation worked well, it remembers that too. The list gets sharper the longer you use it.
Recommendations come in two tiers: small repeatable best-practice tactics that ship in days, and strategic content or off-site campaigns that play out over weeks.
Most current GEO advice collapses these tiers into one undifferentiated list. A brand asks “what should I do?” and gets back a single feed that mixes “add answer capsules to your top three product pages” with “build a category-leading research report and earn citations from twelve industry publications.” Both are real recommendations; they need entirely different ownership, timelines, and proof. Sill separates them so your calendar reflects what your team can actually staff this month, not a wish list that mixes hours of work with weeks of work and asks you to pick.
| Tier | Example | Cadence | Expected Lift |
|---|---|---|---|
| Best practice | Add 19+ statistics to a money page; insert answer capsules under each H2; mark up FAQs with schema | Weekly sprints, 1-2 days per item | 1-3 SOV points per change, compounding |
| Strategic content | Build a comparison page covering five competitors; commission a 2,500-word category report | Two- to four-week initiatives | 5-15 SOV points on targeted query groups |
| Off-site presence | Claim and complete a G2 profile; pitch a podcast in a citation-heavy venue; correct a Wikipedia inaccuracy | Multi-week, depends on third parties | Variable; tracked as a strategic initiative |
| Watchdog remediation | Counter a factually incorrect AI claim about pricing or features with a public source page | Triggered immediately on alert acknowledgment | Defensive: prevents reputation drift |
The expected lift estimates are anchored in the published research on what gets cited by AI engines. Aggarwal et al. (KDD 2024) tested nine optimization methods across 10,000 queries and found that adding statistics produced a 30-40% visibility improvement. Wu et al. (CMU 2025) measured a 35.99% lift from AutoGEO-optimized content. SE Ranking found that pages with 19 or more statistics earn 93% more AI citations and that answer capsules appear in 87% of cited posts. Otterly.AI's controlled experiments confirmed that listicle inclusion and footer text repetition produce measurable lifts while llms.txt and author pages do not. Every recommendation Sill surfaces is paired with the research that made it a priority, so when you ask “why am I being told to do this?” the answer is one click away and you can defend the work to your team and your CFO before you start it.
Mark a recommendation as done and Sill automatically measures whether it moved your visibility on each AI platform, against the prompts that the change should have affected.
The biggest difference between Recommendations and a generic GEO checklist is what happens when you mark something complete. You do not have to set up an experiment, choose a control group, pick a window, or wait for an analyst to run the numbers. Sill already knows which of your prompts the change is most likely to affect, watches what those prompts do on ChatGPT, Gemini, Perplexity, Google AI Overviews, Claude, and Grok, and tells you whether the change made a real difference once enough data is in. You go back to writing the next thing.
The measurement is honest in a way that before-and-after dashboards cannot be. Sill compares the prompts the change should have moved against the prompts it should not have, so a brand-wide visibility bump from a model update does not get falsely credited to your work. Each AI platform is measured separately, so a lift on Perplexity does not get averaged out by flat movement on Claude. As we wrote in the GEO proof gap, naive before-and-after comparisons are the structural weakness of every monitoring tool on the market today; running each completed recommendation through real statistical controls is how Sill closes that gap one action at a time.
Negative results are first-class. If a change made things worse on a platform that matters to you, Sill tells you and surfaces a one-click recommendation to revert it, with the experiment result attached so you do not have to make the case to your team from memory. You get to learn from every action you ship, including the ones that did not work.
Recommendations are scheduled into a four-week calendar that paces to how much work your team actually ships, so you see a sprint you can finish, not a backlog you will not touch.
A list of work with no schedule is a backlog dressed up as a roadmap. The Recommendations calendar gives you a four-week view of what is queued for the upcoming sprint, which longer-running campaigns are in flight, and which experiments will deliver results before your next planning meeting. Related work is batched together so a single sprint focuses on a coherent set of pages and prompts, which makes the visibility movement easier to read once results come back.
Most importantly, the calendar paces to your actual throughput. If your team ships five small changes a week, you see five-item sprints. If your team ships one, you see one-item sprints with more room for the strategic work that takes longer. You do not get handed a 50-item wish list every Monday; you get a sprint your team can realistically finish, plus a clear view of what is coming next. Recommendations should respect your calendar, not pretend it does not exist.
Sill prioritizes the tactics that produced measurable lifts on your brand and your queries, so each sprint builds on what is already working for you instead of restarting from generic best practices.
The published GEO research is a strong starting point, but it cannot tell you which tactics will work on your specific brand, in your category, against the competitors you are actually losing to. Sill closes that gap by learning from the experiments your own work generates. When a tactic produces a measurable visibility lift on your brand, similar recommendations move up in priority for your future sprints. When a tactic falls flat on your brand twice in a row, it moves down. Over time, your queue becomes a portrait of what actually works for you.
The ROI case follows directly from the loop. AI referral traffic converts at 14.2% versus 2.8% for standard organic, a 4.4x premium documented across Microsoft Clarity, Adobe, and Exposure Ninja datasets. A brand that moves visibility by even a few points on a query group worth 1,000 monthly AI sessions can pick up dozens of additional high-intent referrals every month, each worth several times what an equivalent organic referral would be worth. The value of Recommendations is not any single suggestion; it is the discipline of running the actions that worked best on your own data, on a schedule your team can keep, with proof at the end of every cycle.
Recommendations is live in your Sill dashboard now, with a calendar view, a prioritized backlog, a strategic initiatives tracker, and a detail page for every recommendation.
Open Recommendations from your Sill sidebar and you will see the upcoming sprint and the four-week forecast in a calendar view, the full backlog ranked by priority, the strategic initiatives tracker for multi-week campaigns, and a brand strategy editor where you tell Sill how you want to be positioned, which perceptions to lean into, and which competitors you most want to outrank. Every recommendation has its own detail page with the supporting research, a clear description of what to do and how to do it well, and the experiment status once you have completed the action.
If you already have monitoring set up, your first set of recommendations is waiting on your next dashboard refresh. If you are new to Sill, you will see them within a day or two of your first monitoring run completing. There is nothing to install and nothing to configure beyond what you have already set up.
Sill turns your monitoring data into a prioritized, scheduled, measurable plan of work, then verifies each action with statistical controls so the next sprint is built on what actually moved the needle on your brand.
Request your first analysis today to see where you stand.