You decide what to work on. Penlid researches it, writes it, fact-checks every claim, and publishes once you approve the draft.
Then it measures whether AI search actually recommends you — Claude with live web search always, and every other engine this deployment has a key for, sampled repeatedly and reported with real confidence intervals — and ties every movement back to the exact change it follows.
Publishing volume is no longer the thing that pays. These are measured, published figures — not our marketing.
of US Google searches ended without a click in early 2026 — 74% for informational queries.
SparkToro / Similarweb, 2026clicks per article, across 2,000 AI-written pages on 20 fresh domains over 16 months.
SE Ranking experiment, Mar 2026Google search referrals to publishers year on year, measured across 2,500+ sites.
Reuters Institute / Chartbeat, 2026We are not going to pretend otherwise. Generic informational volume on a new domain is, in our view, the weakest content strategy available in 2026. Penlid is built for what still works: refreshing pages that already have equity, publishing commercial-intent content, and being citable by AI engines — then proving which of those actually moved.
Every other tool in this category does one or two of these. The third is the one that changes the conversation with your client or your boss.
Live research against primary sources, a drafted piece in your brand voice, then a fact-verification pass that records every factual claim, its verdict, and the quoted source that supports it. You get an artifact you can hand to a client.
WordPress, a Git repository, your own webhook, or a draft kept here for you to move by hand — with JSON-LD schema, internal links chosen from your real site graph, and a genuine editorial workflow: comments, approval states and version history.
Repeated sampling of Claude with live web search, plus Search Console, reported with Wilson confidence intervals. Every change is timestamped in a ledger and compared against matched control pages — so "it went up" becomes a number you can defend. Our guides show how to measure a change like this yourself.
Every intervention — a refresh, a schema block, six new internal links — is timestamped with a full before/after diff. When impressions or citation share move later, the movement is joined back to the change, and measured against matched control pages using difference-in-differences.
We label association as association. A claim of causation appears only when a matched control set exists.
LLM answers are stochastic — ask twice, get two answers. Every competitor reports a single point estimate as if it were a measurement. Penlid samples repeatedly, reports a Wilson interval, records the model version behind every sample, and will tell you "no significant change" when that is the truth.
Claude, with live web search, is always sampled. More are added as a key and an adapter are configured for each, up to your plan's engine count — the free check names the ones this deployment can sample right now, and every report names the engines that actually answered.
Penlid watches your Search Console daily, separates real decay from seasonality and algorithm noise, and marks the pages losing ground so you can queue the refresh — with a page's own history in the article brief. Refreshing a page with existing equity is, in our view, the best return on content work — and it is what the tools built around "30 new posts a month" cannot do.
Draft, review, comment, approve, publish — with version history, and evidence re-checked the moment you edit a sentence a claim was verified against. Google's spam policies are aimed at pages made at scale to manipulate rankings rather than to help people, so we built the workflow that makes review actually happen.
Comparison against the common feature set of budget autoblogging tools in this category. Verify any row yourself.
| Capability | Typical AI blog tool | Penlid |
|---|---|---|
| Long-form generation, brand voice, multi-language | ✓ | ✓ |
| Auto-publish to WordPress, Git or your own webhook | ✓ | ✓ |
| Internal linking & schema markup | ✓ | ✓ |
| Per-claim fact verification with source records | — | ✓ |
| Editorial workflow: approvals, comments, versions | — | ✓ |
| Search Console integration with a 16-month backfill | — | ✓ |
| Content decay detection | — | ✓ |
| Refresh drafts briefed on the page's own performance history | — | ✓ |
| AI citation tracking with repeated sampling | — | ✓ |
| Confidence intervals on every measurement | — | ✓ |
| Change→effect attribution ledger | — | ✓ |
| Matched-control difference-in-differences | — | ✓ |
| Multi-client workspaces with read-only client logins | — | ✓ |
| REST API with scoped keys | — | ✓ |
Rows reflect the common feature set of budget autoblogging tools as of August 2026. Individual products vary; check yours before switching.
Penlid does not pick your topics or decide which page to refresh — you do, and the dashboard is built to make that call easy. Everything after it runs on its own.
Priced on what actually costs money to run — pages under management and measured prompts — so your bill tracks your value instead of an arbitrary article count.
One site. Generation, fact-checking, publishing, decay detection and AI-visibility measurement, with real confidence intervals.
See individual plansPooled across client workspaces. Read-only client logins scoped to their own brand, per-client health, and what each client costs you to serve — never shown to the client.
See agency plansFree, no account. We ask Claude, with live web search, four real buying questions in your category and show you exactly which competitors get named instead of you.
Costs us real money per run. That's rather the point — it's why a $9 clone can't copy it.