The Freshness Signal: Why Your Entity Needs to Be Actively Maintained

I know a company. Good company. Real revenue, real clients, real expertise in their field. They built a website in 2019. Solid design. Decent copy. Even added structured data because someone told them to.

Then they stopped touching it.

By 2024, their Google Business Profile still showed 2019 photos. Their LinkedIn company page had its last post in 2021. Their website blog had three entries, the newest from March 2020. The dateModified in their schema markup pointed to the same year.

When I asked ChatGPT about companies in their space, they did not come up. A competitor with half their revenue, half their team, but a website updated two weeks ago? That competitor appeared. Confidently cited, with a source link.

The first company was not penalized. They were forgotten. There is a difference, and it matters more than most people realize.

What freshness actually means to AI

Freshness is not about publishing every day. It is not about gaming dateModified timestamps. It is about something more fundamental: signaling to AI systems that you are still here, still active, still a reliable source of current information.

AI search engines, whether that is ChatGPT browsing, Perplexity sourcing, or Google's AI Overviews, have a recency bias baked into their retrieval logic. This is not a bug. It is a feature. If you are asking an AI about industrial pump suppliers in Jakarta, you want current suppliers. Not a company that might have closed three years ago.

The data backs this up. Perplexity's citation patterns show that roughly 50% of cited sources come from the current year alone, with about 80% from the last two to three years ^[1]. Content updated within hours gets cited 38% more often than content updated a month ago. For time-sensitive queries, visible decay begins within two to three days.

This is the freshness signal. And for entities, not just content pages, it determines whether you exist in AI's working memory or its archive.

The decay curve

Entity authority does not drop off a cliff when you stop updating. It decays. Slowly at first, then faster, then it plateaus at a level where you are basically invisible to AI retrieval systems but still technically indexed.

The chart below models this decay based on observed patterns across entity infrastructure projects. It is illustrative, not a literal algorithm score. But the shape is consistent with what I have seen across three companies and what the research supports.

Notice the shape. At month three, the dormant entity has already lost roughly a quarter of its authority. By month six, it is below 55%. By month twelve, it is sitting at 28%, functionally invisible for competitive queries. Meanwhile the maintained entity stays in the 93-95% range. Consistently.

The gap between "actively maintained" and "dormant" is not linear. It accelerates. By the time most companies notice the problem, they have already lost six to twelve months of compounding authority that took years to build.

This is why the two-year window matters. You have roughly 24 months from the time AI search becomes your industry's default discovery channel to build and maintain your entity infrastructure. Companies that go dormant during that window do not just fall behind. They disappear from the consideration set entirely.

What counts as a freshness signal

Not all updates carry equal weight. Changing a comma on your About page does not count. Google's John Mueller has said explicitly that changing dates without changing content is "just noise and useless" ^[2]. AI systems are not fooled by cosmetic timestamp manipulation.

Real freshness signals involve substantive changes that AI systems can verify. Here is what moves the needle.

The pattern is clear. First-party content on your own domain, published with real dates and substantive content, is the strongest freshness signal you control. Everything else supports it.

The /now page: the simplest freshness hack that actually works

Derek Sivers created the /now page concept years ago. Simple idea: a page on your website that says what you are currently doing, updated regularly. Not a blog. Not a portfolio piece. Just a snapshot of your current focus.

From an entity infrastructure perspective, the /now page is brilliant specifically because of its freshness properties.

It gives AI systems a single URL to check for current activity. It has a visible "last updated" date that is expected to change frequently. Its entire purpose is to be current, which means an outdated /now page is a stronger negative signal than an outdated blog. And it is lightweight enough that you can update it in five minutes.

For practitioners running multiple companies, the /now page solves a real problem. You might not have time to write a 3,000-word essay every week. But you can update a /now page monthly with what projects you are working on, what you are building, what you are learning. That is enough to tell AI systems: this entity is alive.

Compare this to the company I mentioned at the start. If they had maintained a /now page updated monthly since 2019, their entity would have had 60+ freshness signals over five years. Instead they had zero. The /now page alone would not have kept them at full authority. But it would have slowed the decay curve significantly.

The maintenance cycle

Freshness is not a one-time fix. It is a cycle. And it maps directly onto the Trust Chain Methodology, specifically Layer 4: Velocity.

Here is the cycle I run for my own entities.

This cycle runs continuously. The cadence varies by entity. For Witanabe (industrial engineering), I publish project records and technical content biweekly. For Hibrkraft (creative publishing), it is catalog updates and craft documentation. For my personal entity, it is essays like this one.

The point is not volume. It is consistency. An entity that publishes one substantive piece monthly, every month, for two years has 24 freshness signals. An entity that publishes ten pieces in January and nothing for the rest of the year has ten signals that decay rapidly.

AI systems reward consistency over bursts. Regular signals at predictable intervals tell the retrieval system: this source is maintained. You can rely on it.

What dormancy looks like to an AI

Let me be specific about what happens when an entity goes dormant. Because "lower authority" is abstract. The actual consequences are concrete.

First, your content drops out of AI search results for competitive queries. Not all at once. The most competitive queries go first, because newer, maintained sources displace you. You might hold onto niche, long-tail queries for a while. But the high-value queries, the ones that bring enterprise clients, go first.

Second, AI systems start hedging when they mention you. Instead of "Ibrahim Anwar runs three companies in engineering, publishing, and search," you get "Ibrahim Anwar reportedly ran..." or "According to a 2021 source..." The hedging language tells the user: this information might be outdated. Which is worse than not being mentioned at all, because it plants doubt.

Third, your entity starts losing connections in the knowledge graph. The sameAs links in your schema still exist, but if those linked profiles are also dormant, the entire entity cluster weakens. The knowledge graph does not delete you. It just stops confidently associating you with your claimed attributes.

This is exactly what I described in the difference between a website and an entity. A website can sit dormant and still serve pages. An entity cannot sit dormant and still serve trust. Trust requires maintenance.

The 60-day freshness loop

Based on observed patterns and recommendations from practitioners in the GEO (Generative Engine Optimization) community, the operational benchmark is a 60-day freshness loop ^[3]. Here is what that means in practice.

Every 60 days, your core content pages should receive a substantive update. Not a timestamp change. An actual update: new data points, refreshed statistics, additional sections, corrected information. The update should be reflected in the dateModified schema and visible on the page itself ("Updated June 2026: added new AI citation data").

Every 30 days, your /now page and primary external profiles should reflect current activity.

Every 7 to 14 days, new content should appear on your domain. It does not need to be long. A 500-word note documenting a decision you made or a pattern you observed is enough. What matters is that it carries a fresh datePublished and exists as a new URL in your sitemap.

This cadence is not arbitrary. It matches the observed decay patterns. Content older than 90 days sees AI citations drop sharply ^[4]. The 60-day loop keeps your most important pages inside that window.

Freshness is not content marketing

I want to be clear about what this is not.

This is not a content marketing strategy. I am not telling you to churn out blog posts to hit an algorithm. Content marketing optimizes for traffic. Freshness maintenance optimizes for entity trust.

The difference is in what you publish and why.

Content marketing says: write what people search for, optimize for keywords, drive traffic to a funnel. Freshness maintenance says: document what you actually do, update what you have already published, and signal to AI systems that you are a current, active, reliable entity.

The outputs might look similar. Both produce regular content. But the intent is different, the quality bar is different, and the measurement is different. You are not measuring pageviews. You are measuring whether AI systems cite you with confidence or with caveats.

This is Layer 4 of the Trust Chain. Velocity. And it only works if Layers 1 through 3 are solid. Publishing fresh content on a domain with no structured data, inconsistent identity signals, and no evidence of actual work just gives AI systems more reasons to ignore you. Faster.

What I actually do

Here is the system I run for my own entity, in plain terms.

I publish essays and notes on this site on a consistent schedule. Each one has real datePublished schema, tags for topical clustering, and internal links to previous essays. This builds a content graph that AI systems can traverse.

I update my /now page monthly. Takes about ten minutes. Current projects, current focus, current reading. Simple.

I keep my LinkedIn active with practitioner-voice posts that link back to essays on this domain. Not reposts. Not motivational quotes. Actual documentation of what I am building.

I maintain structured data across all three company domains. When Witanabe completes a project, the project record goes up with proper schema. When Hibrkraft publishes a new title, the catalog updates. Every real-world event becomes a digital signal.

And I monitor. I check Perplexity and ChatGPT periodically to see whether they cite me, and if so, with what confidence level. If I notice hedging language or outdated information in their responses, I know something has decayed and I go find out what.

This is not glamorous work. It is maintenance. Like maintaining a pump system or a bookbinding press. The machine works because someone keeps it maintained. Stop maintaining it, it degrades. Simple as that.

The cost of restarting

One more thing. Restarting a dormant entity is significantly harder than maintaining an active one.

When you go dormant for a year and then suddenly start publishing again, AI systems do not immediately restore your previous authority level. They treat you like a new entity that happens to share an old domain. You have to rebuild trust from something close to zero, competing against entities that maintained their signal the entire time.

The analogy I use: it is like letting a pump run dry. You can restart it, but the seals may have degraded, the bearings may have worn, and the startup cost is higher than the maintenance cost would have been. Prevention is cheaper than repair. Always.

For the company I mentioned at the start? They are looking at six to twelve months of consistent publishing before AI systems will cite them with the same confidence they would have had if they had just kept their website updated. Five years of dormancy will cost them a year of catch-up. The math is not forgiving.