Helpful Content Update

TL;DR — The Helpful Content UpdateThe Helpful Content Update (HCU) was a series of Google updates starting in August 2022 that added a site-wide, machine-learning classifier to demote content made primarily to rank rather than to help people. In March 2024 it was folded into Google's core ranking system. (HCU) was Google’s way of demoting pages made mainly to rank in search instead of to help real people. The twist that surprised everyone: it judges your whole site, not just individual pages — so a pile of thin contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. can drag down even your good pages. It started in August 2022 and, since March 2024, it’s baked into Google’s normal ranking, so there’s no standalone “HCU” anymore.

What the Helpful Content Update is

Google introduced the helpful content system to reward content made primarily for people rather than content made mainly to attract search visits. Evidence for this claim Google's people-first content guidance asks whether content is created primarily to help an intended audience rather than manipulate search traffic. Scope: Google Search self-assessment guidance; it is not an exposed per-page score or exhaustive ranking checklist. Confidence: high · Verified: 2026-07-14 Google: Creating helpful, reliable, people-first content Google later incorporated that system’s signals into its core ranking systems, so it is no longer a separately announced periodic update. Evidence for this claim Google says the former helpful content system became part of its core ranking systems in March 2024. Scope: Google's ranking-systems documentation; core systems use many signals and do not expose a single helpful-content metric. Confidence: high · Verified: 2026-07-14 Google: A guide to Search ranking systems

The Helpful Content Update — most people just say HCU — was a series of Google updates aimed at one idea: reward people-first content and demote search-engine-first content.

• People-first content is made because you have an audience that will find it useful — it shows real, first-hand knowledge and leaves the reader feeling like they actually learned something.
• Search-engine-first content is made mainly to rank: chasing trending keywords, hitting a word count, mass-producing pages on topics you don’t really know, or summarizing what everyone else already said without adding anything.

The one thing most people get wrong

HCU is a site-wide signal. It doesn’t pick out one bad page and dock it. Instead it builds a picture of your entire site’s quality. If a big chunk of your site looks made-for-search-engines, Google can suppress the whole domain — including the genuinely helpful pages you’re proud of.

That’s why “I’ll just fix my five worst pages” rarely works. The classifier is looking at the overall character of the site.

Is it still a thing?

Sort of — but not as a separate update. In March 2024, Google folded the helpful content system into its core ranking algorithm. So if you’re searching for “the latest helpful content update,” there isn’t one. The idea didn’t go away; it just became an always-on part of normal ranking. There won’t be another standalone HCU.

Does it ban AI content?

No. This is the single most common myth. Google has been clear that how content is produced — by a person or by AI — isn’t the issue. What matters is whether it’s original, helpful, and made for people. Plenty of AI-assisted content ranks fine. Plenty of human-written content is search-engine-first junk.

Want the deeper version — how the classifier actually works, the full timeline, why programmatic sites got hit hardest, and what recovery really looks like? Switch to the Advanced tab.

TL;DR — HCU was a series of updates (Aug 2022 → Mar 2024) that added a site-wide, machine-learning classifier to demote content made primarily to rank rather than to help people. It’s a domain-level signal — even your best pages can be suppressed if the site overall reads as search-engine-first — and it runs continuously. It hit programmatic and templated sites disproportionately, and in March 2024 it was absorbed into core ranking alongside the new scaled-content-abuse spam policy, so there are no standalone HCU updates anymore. Recovery is slow and often incomplete; community data shows most hit sites never fully came back. Production method (AI vs human) was never the criterion — intent and value are.

What it actually is

Helpful-content guidance describes a sitewide quality objective and self-assessment framework, not a score exposed in Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.. Evidence for this claim Google's people-first content guidance asks whether content is created primarily to help an intended audience rather than manipulate search traffic. Scope: Google Search self-assessment guidance; it is not an exposed per-page score or exhaustive ranking checklist. Confidence: high · Verified: 2026-07-14 Google: Creating helpful, reliable, people-first content Since March 2024, Google has described the former helpful content system as part of its core ranking systems. Evidence for this claim Google says the former helpful content system became part of its core ranking systems in March 2024. Scope: Google's ranking-systems documentation; core systems use many signals and do not expose a single helpful-content metric. Confidence: high · Verified: 2026-07-14 Google: A guide to Search ranking systems

The Helpful Content UpdateThe Helpful Content Update (HCU) was a series of Google updates starting in August 2022 that added a site-wide, machine-learning classifier to demote content made primarily to rank rather than to help people. In March 2024 it was folded into Google's core ranking system. introduced a site-wide classifier: a continuously running machine-learning model that scores how much of a site looks like it exists to rank in search engines versus to genuinely help an audience. The more search-engine-first content a site carries, the stronger the negative signal it gets — it’s weighted, not a binary on/off switch.

The defining property, and the source of most confusion in the industry, is that it’s domain-level, not page-level. Google said it directly: “Any content — not just unhelpful content — on sites determined to have relatively high amounts of search-engine-first content is less likely to perform well in Search, assuming there is other content on the web that is better to display.” If you remember one sentence from this page, make it that one. It’s why a site with 10% excellent pages and 90% thin pages sees the excellent pages drop too.

If you’ve been doing this a while, the mental model is Panda redux — a site-level quality signal that propagates down to the page level. As Google’s Paul Haahr put it, “we’re propagating information from the domain to the page level.”

How it differs from a core update

A classic core update is a broad, periodic reassessment of relevance and quality across all of search. HCU was narrower and more specific: a dedicated classifier targeting the people-first/search-engine-first distinction, running continuously rather than on a refresh cycle. After March 2024 that distinction collapsed — the helpful content system is now part of core ranking, so the two are no longer separate things.

The timeline

One thing worth getting right: there was no standalone HCU in August 2023 — that was the August 2023 broad core update. The September 2023 HCU followed it about a week later.

How the classifier works

A few technical points that competitors usually skip:

• It’s machine learning, not manual review. No human rater is grading your pages one by one; it’s an automated model evaluating site-level signals.
• It runs continuously. Unlike a periodic refresh, the classifier is always on. Google: “If it sees a site has reduced unhelpful content, the site might start performing better at any time.” In theory. In practice, see the recovery section.
• It’s weighted. Sites with more unhelpful content get a stronger negative signal — Danny Sullivan: “Sites with a lot of unhelpful content will find the signal stronger for them.”
• Third-party and subdomain content counts. The September 2023 update clarified that low-quality contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. hosted on your subdomains, or third-party “parasite” content on your domain, is assessed as part of your site and can drag the whole host down.

Why programmatic SEO got hit hardest

This is the part that matters most for this pillar. HCU disproportionately punished programmatic and templated sites, and the reason is structural.

Programmatic SEOProgrammatic SEO (pSEO) is the practice of generating many pages from a single template plus a data source to target large sets of similar queries. It's powerful when each page genuinely answers its query with unique data, and spam when it just stamps a thin template across a shallow dataset. produces pages at scale across many topics or locations from one template plus a dataset. When the dataset is shallow, the output is exactly what HCU was built to demote: mass-produced, automated, search-engine-first pages whose only differentiator is a swapped variable (Best plumbers in [City]). My standing line on this applies directly — thin contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. at scale is a data problem, not a template problem. If deleting the modifier leaves a generic page, the classifier has nothing to reward.

March 2024 made the connection explicit with the new scaled-content-abuse policy: producing content at scale primarily to rank, with little unique value per page, “no matter how it’s created.” That’s a near-perfect description of low-value pSEO.

Worth keeping two different things separate here, because they get conflated constantly. The helpful-content effect on core ranking is algorithmic — no notice, no manual action, no reconsideration request, just lower visibility as the ranking systems assess the site. Scaled content abuseScaled content abuse is Google's spam policy (introduced March 2024) for generating many low-value pages primarily to manipulate search rankings rather than help users — and it applies no matter how the content is created: AI, automation, or human writers. is a distinct, named spam policy that Google enforces “both through automated systems and, as needed, human review that can result in a manual action” — the kind that shows up in Search ConsoleGoogle's free tool for monitoring crawling, indexing, and search performance.’s Manual Actions report, with a path to reconsideration once fixed. A site can have a genuine quality problem with no manual action attached, or a manual action for scaled content abuseScaled content abuse is Google's spam policy (introduced March 2024) for generating many low-value pages primarily to manipulate search rankings rather than help users — and it applies no matter how the content is created: AI, automation, or human writers. specifically. Related levers, not the same lever.

The damage to programmatic and review-aggregator sites was real and, by most accounts, large — but I don’t have a single clean, Google-published number for “how much,” and I’m not going to invent one. Community trackers describe steep, sometimes near-total declines for the thinnest sites; what’s consistent across that reporting is that the sites with nothing but a swapped variable lost the most. The dividing line between surviving and penalized pSEO is the same as it always was — unique data per page, real expertise, and a reason to exist beyond the ranking.

Google’s self-assessment questions

Google published a set of questions to evaluate your own content. They’re worth running honestly. A representative cut:

Content & quality

• Does your content provide original information, reporting, research, or analysis?
• Does it give a substantial, complete, or comprehensive description of the topic?
• Does it provide insightful analysis or interesting information beyond the obvious?
• If it draws on other sources, does it add substantial value and originality rather than just copying or rewriting them?
• Is this the sort of page you’d want to bookmark, share, or recommend?

Expertise

• Does the content make you want to trust it — clear sourcing, evidence of expertise, author/site background?
• Would researching the site leave the impression it’s a well-trusted authority on its topic?
• Is it written by an expert or enthusiast who demonstrably knows the topic?

Presentation & production

• Is it produced well, or does it appear sloppy or hastily made?
• Is it mass-produced or spread across a large network of sites so individual pages get little care?
• Do excessive ads interfere with the main content?

And the search-engine-first red flags to avoid: writing for keywords you saw others rank for, “extensive automation to produce content on many topics,” chasing a word count because you think Google prefers one (it doesn’t), and artificially “freshening” content when nothing substantive changed.

Recovery: what Google says vs. what actually happened

Here’s where I’ll be blunt, because the honest version isn’t the comfortable one.

What Google says: recovery is possible at any time — reduce unhelpful content, demonstrate sustained improvement over a period of months, and the always-running classifier can re-evaluate you. John Mueller: it “can take months (6+ months) for Google to reevaluate a site after improving quality overall.”

What actually happened: recovery has been slow and frequently incomplete.

• A community study tracking 200+ sites foundA 302 (\"Found\") is a temporary redirect: it forwards users to a new URL while telling search engines the original URL should stay in the index. It's a weak canonicalization signal, not the zero-equity dead end of SEO folklore. zero recoveries from the September 2023 HCU as of the March 2024 core update — roughly seven months with nothing.
• After the August 2024 core update, some sites finally moved: roughly 22% of tracked sites (~88 of ~400) saw 20%+ traffic increases. Better than zero, but a minority, and “up 20%” is not “back to where I was.”
• Danny Sullivan acknowledged it directly: the changes “have helped some of those sites but generally have not brought those sites all the way back up to the level they were.”

Why it’s so slow and uncertain:

1. The classifier wants to see sustained improvement, not a one-time content sprint — there’s effectively a probation period.
2. Recovery has tended to come in waves tied to major core updates, not continuously.
3. The ranking landscape itself shifted underneath everyone. Sullivan again: “September doesn’t exist anymore. And our ranking systems are different.” Even a perfect site may not get the old rankings back, because the competitive set and the systems both moved.

And the most important reframe, from Mueller: these “are not ‘recoveries’ in the sense that someone fixes a technical issue and they’re back on track — they are essentially changes in a business’s priorities.” That’s the truth nobody hit by this wants to hear. There is no tag to add, no noindex sprint, no AI rewrite that fixes it. It’s a change in what the site is for.

Common myths, quickly

• “HCU banned AI content.” No. Production method isn’t the criterion — the September 2023 wording change from “by people” to “created for people” removed the by-a-human requirement entirely. Original, helpful AI-assisted content is fine.
• “It targets a word count.” No. Targeting a word count “because you believe Google has a preferred word count” is itself listed as a search-engine-first signal.
• “It only affects your bad pages.” No — it’s site-wide; good pages get suppressed too.
• “Recovery is fast — fix and wait a few weeks.” No. Months at minimum, often longer, often incomplete.
• “Deleting thin pages guarantees recovery.” Necessary, not sufficient — and mass deletions can create their own crawl and internal-link problems.
• “There’s a new HCU coming.” No. Since March 2024 it’s part of core ranking. There are no more standalone helpful content updates.

What about Bing?

Microsoft never shipped a named “helpful content update,” but its guidance landed in the same place: people-first over search-engine-first, with weight on depth, completeness, and E-E-A-T, plus engagement signals (dwell time, bounce-back). Bing has been even more explicit than Google that AI/automation isn’t itself the problem — “use of AI or automation is not against Bing’s guidelines, as long as it is not primarily used to manipulate search rankings.”

Bottom line

HCU formalized something that was always true: Google is trying to surface content made for people, and it now judges that at the level of the whole site, continuously, as part of core ranking. If your site is a vehicle for ranking, that’s the thing the last few years of updates were built to identify. If it’s a genuinely useful resource — with real expertise and data only you have — you were never the target. Recovery, if you’re hit, is a business-priorities change, not a patch.

AI summary

A condensed take on the Advanced version:

• HCU = a site-wide, machine-learning classifier that demotes search-engine-first content in favor of people-first content. Weighted, not binary; runs continuously.
• Domain-level, not page-level. Even your best pages can be suppressed if the site overall looks made-to-rank — “any content — not just unhelpful content” can drop. It’s effectively Panda-style: domain signal propagated to the page level.
• Timeline: Aug 2022 (English launch) → Dec 2022 (global) → Sep 2023 (the big, most impactful one; language shifted from “by people” to “for people”) → Mar 2024 (folded into core ranking; new scaled-content-abuse + expired-domain-abuse policies; Google expected a 40% cut in unhelpful content and later reported 45%). No standalone HCU exists anymore.
• Core-ranking effect vs. spam policy: helpful-content’s effect on ranking is algorithmic (no notice, no manual action); scaled content abuseScaled content abuse is Google's spam policy (introduced March 2024) for generating many low-value pages primarily to manipulate search rankings rather than help users — and it applies no matter how the content is created: AI, automation, or human writers. is a separate spam policy that can trigger an actual manual action in Search ConsoleA free Google service that reports how a site performs in Google Search and surfaces problems with how Google crawls, indexes, and serves it. It's first-party data straight from Google — but you don't need it to appear in results.. Related, not the same lever.
• Programmatic SEOProgrammatic SEO (pSEO) is the practice of generating many pages from a single template plus a data source to target large sets of similar queries. It's powerful when each page genuinely answers its query with unique data, and spam when it just stamps a thin template across a shallow dataset. got hit hardest: templated pages where only a modifier changes are exactly what it demotes. Thin contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. at scale is a data problem, not a template problem. Community trackers describe steep declines for the thinnest sites; there’s no single Google-published figure for “how much,” so none is invented here.
• Recovery is slow and often incomplete. ~Zero recoveries in the first 7 months after Sep 2023; ~22% of tracked sites up 20%+ after Aug 2024. Sullivan: most sites “not… all the way back.” Mueller: it’s “changes in a business’s priorities,” not a technical fix.
• Myths busted: no AI ban; no word-count target; not page-only; not fast; deletion alone isn’t enough; no new standalone HCU coming.
• Bing converged on the same view: people-first, value over method — AI isn’t the problem unless used to manipulate rankings.

Official documentation

Primary-source guidance from the search engines on helpful content.

Google

• Creating helpful, reliable, people-first content — the self-assessment questions, the people-first vs. search-engine-first framing, and the “Who, How, Why” guidance.
• Helpful Content Update (Aug 2022 announcement) — the original launch post introducing the site-wide signal.
• What web creators should know about our March 2024 core update and new spam policies — the update that absorbed the helpful content system into core ranking and launched the new spam policies.
• Spam policies — Scaled content abuse — the March 2024 policy most relevant to programmatic sites: pages made to rank, “no matter how it’s created.”
• Google Search’s guidance about AI-generated content (Feb 2023) — production method isn’t the criterion; quality and intent are.

Bing / Microsoft

• Bing Webmaster Guidelines — content quality — Bing’s people-first framing and its position that AI/automation isn’t itself against the guidelines unless used to manipulate rankings.

Quotes from the source

On-the-record statements from Google about the helpful content system, the site-wide signal, and recovery.

Google — the site-wide signal

• “Any content — not just unhelpful content — on sites determined to have relatively high amounts of search-engine-first content is less likely to perform well in Search, assuming there is other content on the web that is better to display.” — Google Search Central.
• “Sites with a lot of unhelpful content will find the signal stronger for them.” — Danny Sullivan, Google Search Liaison.
• “Google looks at the website overall, so if it sees a lot of low-quality contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count., then Google can take that into account for rankings.” — John Mueller, Google.
• “We’re propagating information from the domain to the page level.” — Paul Haahr, Google ranking engineer.

Google — recovery

• “The classifier is always running. If it sees a site has reduced unhelpful content, the site might start performing better at any time.” — Danny Sullivan, Google.
• “The changes have helped some of those sites but generally have not brought those sites all the way back up to the level they were.” — Danny Sullivan, Google.
• “September doesn’t exist anymore. And our ranking systems are different, and among other things, our ranking systems are also rewarding other kind of content too.” — Danny Sullivan, Google.
• “I wish we could do a better job by those sites as well — these are really good things in our systems need to improve.” — Danny Sullivan, Google.
• “It can take months (6+ months) for Google to reevaluate a site after improving quality overall.” — John Mueller, Google.
• “These are not ‘recoveries’ in the sense that someone fixes a technical issue and they’re back on track — they are essentially changes in a business’s priorities.” — John Mueller, Google.

Google — content philosophy

• “Create your content for people, not robots, for success with Google Search.” — Danny Sullivan, Google.

The mental models

1. Site-level, not page-level — the Panda mindset. HCU scores the whole domain and propagates that down to pages. So the question is never “is this page good?” — it’s “what does the overall content posture of this site say about why it exists?” Fixing five pages on a site full of thin contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. changes nothing.

2. The people-first / search-engine-first binary. Every page sits on one side of a line: made for an audience, or made for a ranking. The honest test: would you have made this page if search engines didn’t exist? If the only reason it exists is a keyword you saw someone else rank for, it’s on the wrong side.

3. Thin contentThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. at scale is a data problem, not a template problem. This is the bridge to programmatic SEOProgrammatic SEO (pSEO) is the practice of generating many pages from a single template plus a data source to target large sets of similar queries. It's powerful when each page genuinely answers its query with unique data, and spam when it just stamps a thin template across a shallow dataset.. When pSEO output gets hit, people try to “beef up” each page with more words. Wrong fix. Delete the modifier from any page; if what’s left is generic, the dataset is too shallow. No template polish saves a page with nothing unique to say.

4. Recovery is a priorities change, not a patch. Reframe it the way Mueller does: not a technical fix you ship and revert, but a sustained change in what the business is trying to do. The classifier wants months of evidence that the change is real. Plan for that timeline, or don’t start.

5. Method is not the signal. AI vs. human is the wrong axis. The axes that matter are value, intent, originality, and expertise. Google demoted the language requiring content be “written by people” for a reason — they’re judging the output, not the process.

HCU at a glance — cheat sheet

The facts that get misremembered

Myth → reality, fast

• “Banned AI” → method-neutral; quality and intent decide.
• “Just fix the bad pages” → site-wide; good pages drop too.
• “Fix and wait a few weeks” → 6+ months, and many sites never fully recover.
• “Delete thin pagesThin content is web content that provides little or no value to users. Google's spam policies name it 'thin content with little or no added value' — and it's about value per page, not word count. = recovery” → necessary, not sufficient.
• “A new HCU is coming” → it’s now part of core; no standalone updates.

The programmatic connection

• Templated pages where only a modifier changes = textbook search-engine-first.
• March 2024’s scaled-content-abuse policy targets exactly this — “no matter how it’s created” — and can trigger a manual action, separate from the algorithmic core-ranking effect.
• Survivors had unique per-page data + real expertise; there’s no single Google-published figure for how steep the losses were on the rest, so this page doesn’t invent one.

People-first content self-assessment

Run this on representative pages and on the site’s overall publishing pattern. A single checked box does not prove quality; the point is to expose whether the site consistently serves an audience or mainly manufactures search inventory.

Content and quality

• The page contains original information, reporting, research, or analysis rather than a rewrite of what already ranks.
• It gives a substantial and complete treatment appropriate to the question, without padding toward a supposed preferred word count.
• It adds useful analysis or information beyond the obvious.
• When it uses other sources, it adds substantial value instead of copying or lightly paraphrasing them.
• It is a page the intended audience could reasonably bookmark, share, or recommend.

Expertise and trust

• Claims have clear sourcing where readers would need evidence.
• The author or site background makes the relevant expertise visible.
• First-hand knowledge, original data, or demonstrable experience supports the page where the topic calls for it.
• A reader researching the publisher would understand who is responsible for the content and why they are credible on this subject.

Presentation and production

• The page is carefully produced rather than sloppy, rushed, or internally contradictory.
• Mass production has not reduced the care given to each page or template.
• Ads and interface elements do not obstruct the main content.
• Any automation has editorial oversight for usefulness, accuracy, and originality.

Site-level intent

• The site has an existing or intended audience beyond people arriving from search.
• You would still have a reason to publish this material if search engines did not exist.
• Topics come from the audience and the site’s expertise, not merely from keywords competitors rank for.
• Dates change only when content changes substantively; there is no artificial “freshening.”
• Thin programmatic variants are improved with real data, consolidated, or kept out of the indexStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed. instead of being padded with generic text.

If the site was affected

• The review covers the overall content posture, not just a handful of losing pages.
• The plan represents a sustained change in publishing priorities, not a one-time technical patch or mass AI rewrite.
• Stakeholders understand that reassessment can take months and recovery may be incomplete; no fixed restoration date or ranking guarantee is being promised.

Prompts for a people-first content audit

Review one page without turning the output into a score

Paste the page text, author information, cited sources, and a short description of the intended audience. Expect an evidence table and questions for a human reviewer.

Review this page against Google's people-first self-assessment themes: original and
complete content, added value beyond sources, visible expertise and trust, careful
production, a clear intended audience, and a purpose beyond attracting search visits.

Return:
1. Evidence present in the supplied page for each theme.
2. Missing evidence or unanswered reader needs.
3. Search-engine-first warning signs, including keyword-chasing, artificial freshness,
   unsupported expertise, word-count padding, or automation without adequate review.
4. Questions a subject-matter expert must answer before revision.
5. A recommendation: retain, improve, consolidate, or remove/noindex, with reasoning.

Do not invent expertise, sources, user needs, or a numeric “helpfulness score.” Do not
rewrite the page until the missing facts have been supplied.

[PASTE AUDIENCE, PAGE, AUTHOR DETAILS, AND SOURCES]

Audit a site’s publishing pattern

Paste a content inventory with URL, template or topic cluster, author, source type, publication/update dates, and a short value summary. Expect pattern-level findings.

Audit this content inventory for site-wide people-first versus search-engine-first
patterns. Group pages by template and topic cluster. Flag clusters where pages appear
mass-produced, derivative, outside the site's demonstrated expertise, artificially
freshened, or differentiated only by a keyword modifier.

For each cluster, identify the supplied evidence, what must be manually sampled, and the
most defensible action: keep, improve with original expertise/data, consolidate, or keep
out of the index. Prioritize business-level changes to the publishing model over isolated
technical fixes. State clearly what the inventory cannot prove and do not promise a
recovery date or restored rankings.

[PASTE CONTENT INVENTORY]

Test yourself: the Helpful Content Update

Five quick questions on what HCU is, how it works, and what recovery actually looks like. Pick an answer for each, then check.

Resources worth your time

My related writing

• Programmatic SEOProgrammatic SEO (pSEO) is the practice of generating many pages from a single template plus a data source to target large sets of similar queries. It's powerful when each page genuinely answers its query with unique data, and spam when it just stamps a thin template across a shallow dataset. — my full take on when scaled pages work and when they’re the thing HCU was built to demote.
• What is quality content? — why real expertise and unique data are the differentiators, and why faked expertise fails at scale.
• How to Create Great Content for Search (Ahrefs) — the people-first version of content creation in practice.
• 96.55% of Pages Get No Traffic From Google — the base-rate problem behind Google’s push to surface only genuinely helpful results.

My speaking

• How Search Works (SlideShare) — my walkthrough of crawlingCrawling is how search engines use automated bots (like Googlebot and Bingbot) to discover URLs and download pages. A page has to be crawlable to be indexed, but crawling on its own isn't a ranking factor., renderingTurning HTML, CSS, and JavaScript into the final visual page and DOM., indexingStoring a crawled page in the search index so it can appear in results. Crawled is not the same as indexed — Google selects what to keep, and indexing isn't guaranteed., and ranking, with my standing disclaimer: “This is my understanding of systems… not going to be 100% complete or accurate.”

From around the industry

• Google’s Helpful Content Update Is Sitewide (Search Engine Roundtable) — the clearest early explainer of the domain-level signal.
• Google’s Helpful Content Update Introduces A New Site-Wide Ranking Signal (Glenn Gabe, GSQI) — deep documentation of the site-level mechanism.
• Impact of the September 2023 Helpful Content Update (Search Engine Land) — why the September 2023 update was the most consequential.
• Recovery Uncertain: Danny Sullivan On Algorithm Impacts (Search Engine Land) — the source for the “September doesn’t exist anymore” framing.
• 22% Of Sites Hit By The Helpful Content Update Are Up 20% Or More (Search Engine Roundtable) — the recovery-rate data after the August 2024 core update.
• SEranking: HCU Recovery — Sites Getting Back On Track — tracked-site recovery analysis with named examples.
• Google’s Spam Policies — Scaled Content Abuse — the March 2024 policy that codified the programmatic-SEO connection.

Helpful Content Update

The Helpful Content Update (HCU) was a series of Google updates starting in August 2022 that added a site-wide, machine-learning classifier to demote content made primarily to rank rather than to help people. In March 2024 it was folded into Google's core ranking system.

Related: Programmatic SEO, Indexing

Helpful Content Update

The Helpful Content Update (HCU) was a series of Google ranking updates that began in August 2022 and introduced a site-wide classifier — a continuously running machine-learning model designed to demote “search-engine-first” content (made primarily to rank) in favor of “people-first” content (made primarily to help an audience).

The key thing that made HCU different from a normal core update: it was a domain-level signal, not a page-level one. A high proportion of unhelpful content on a site could suppress the visibility of the entire domain — including its genuinely good pages. Google put it plainly: “Any content — not just unhelpful content — on sites determined to have relatively high amounts of search-engine-first content is less likely to perform well in Search.”

The rollout timeline:

• August 2022 — first launch, English-language searches only.
• December 2022 — expanded globally to all languages.
• September 2023 — the most impactful update; improved the classifier and broadened detection of unoriginal, “made-to-rank” content.
• March 2024 — the helpful content system was absorbed into Google’s core ranking algorithm, ending standalone HCU updates, alongside a new “scaled content abuseScaled content abuse is Google's spam policy (introduced March 2024) for generating many low-value pages primarily to manipulate search rankings rather than help users — and it applies no matter how the content is created: AI, automation, or human writers.” spam policy.

HCU hit templated and programmatic sites especially hard, and recovery has proven slow and uncertain — community tracking shows most affected sites never fully returned to their previous traffic levels. Since March 2024 there are no more standalone “helpful content updates”; the signal is now always-on inside core ranking.

Related: Programmatic SEO, Indexing

Changelog