How to Audit Your Website After a Google Spam Update

After a Google spam update, audit before you act. Confirm your drop actually lines up with the rollout window, then run five passes: freeze the site's current state, compare real content changes against date changes, inspect text readers can't see, check every fact that can go stale, and measure whether each templated page earns its place. Then make one reversible fix, get it live, and measure the same pages again. Deleting half your site on day one destroys the evidence and often the recovery.

What the August 2026 spam update did

Google's August 2026 spam update ran from 18 August to 21 August 2026. Google's Search Status Dashboard logged the start at roughly 12:30pm ET on the Tuesday, and Google confirmed the rollout complete as of 21 August. Google called it "a normal spam update", applying globally and to all languages.

Two details from Google matter more than the announcement. Barry Schwartz at Search Engine Roundtable reported that Google said this update does not target link spam and does not target the site reputation abuse policy. And Google said it will run periodic refreshes of the update, which means a site that was missed in August can still be caught later, and a site that cleans up can be reassessed.

Google would not say what percentage of queries the update affected. Nobody outside Google knows how big it was. Treat any vendor publishing exact impact numbers with suspicion.

This was the third spam update of 2026, two months after the June update. The cadence is accelerating and the rollouts are getting faster: the August 2025 spam update took around four weeks to roll out, and this one took three days. Cheap AI generation has made low-value sites nearly free to produce, and Google is shipping enforcement more often in response. Plan for the next one as routine weather, since one is likely within months.

What a Google spam update is, and what it is not

A spam update is a notable improvement to Google's automated spam-detection systems, the largest of which is SpamBrain, Google's AI-based spam-prevention system. Those systems run continuously. The named update is the moment Google ships a meaningful upgrade to them and logs it on its ranking updates history page.

A spam update is not a core update. A core update reassesses quality and relevance broadly, and you can lose visibility in one without violating anything. A spam update is enforcement. It targets sites that violate Google's published spam policies: scaled content abuse, hidden text and links, doorway pages, expired domain abuse, cloaking, keyword stuffing and the rest. Sites that violate those policies "may rank lower in results or not appear in results at all", in Google's words.

The part that catches site owners off guard: there is no notification. A manual action shows up in Search Console with a message. An algorithmic spam action just moves your rankings. If you were hit, the only place that says so is your own data.

Read the data before you touch the site

I built Radiant Roof Repairs from zero to more than A$1 million a year in revenue using SEO as the only marketing channel. When SEO is your only channel, every update window puts your revenue on the line, and you learn quickly that the worst response to a scary chart is a fast edit. The correct response is a diagnosis.

Start with three checks, in order.

Confirm the timing. Pull Search Console performance data for the weeks before, during and after 18 to 21 August. A drop that started on 12 August is not this update. A drop that starts inside the window and holds after it is a candidate.

Check for manual actions anyway. Spam updates are automated, but the manual actions panel in Search Console costs nothing to check, and it changes everything if something is there. A manual action has its own process: fix the violation, then file a reconsideration request. An algorithmic hit has no form to fill in.

Classify the loss. Nearly every panicked site owner skips this step, and it decides what you do next:

Overlay known events on the same timeline: releases, tracking changes, seasonality, outages. An update window is a comparison window, and the evidence still has to establish the cause. The checklists circulating after every update skip that discipline, and skipping it is how sites end up deleting pages that were never the problem.

The five-pass audit

If the data says the site is the problem, audit it in five passes. Each pass maps to a spam policy Google actually publishes, and each one ends in a finding you can act on, not a feeling.

Pass 1: freeze the baseline

Record the current state before you change anything: the repository commit or a full copy of the site, the route inventory and sitemap, rendered HTML for the affected pages, and the exact Search Console and GA4 windows you're comparing. Every finding that follows attaches to this baseline.

You cannot measure a recovery without a frozen starting point. And if a change makes things worse, the baseline is your way back.

Pass 2: compare the real content diff

Deceptive freshness is the easiest self-inflicted wound to find. If a page's dateModified changed while the visible content stayed the same, you've got a freshness signal with nothing behind it. Scheduled jobs that bump dates across a site are a classic source.

Read the actual diff for the pages that moved. Did the body, facts, examples or links change, or just the date? The repair is honest: make the real update the page needed, then let the date follow the work. A date field alone is not an update.

Pass 3: inspect text readers can't see

Google's hidden text policy covers content placed "solely to manipulate search engines and not to be easily viewable by human visitors": white text on white backgrounds, CSS that positions text off-screen, zero font size, zero opacity.

Check the rendered page as a reader. Then check the CSS states, accessibility labels, image-only claims and JSON-LD. Structured data should describe what's visibly on the page, and a schema block that claims things the page never shows is a risk.

The policy has an explicit carve-out that matters for modern sites: accordions, tabs and other honest show-hide interface elements don't violate it. The test is purpose. A collapsed FAQ a reader can open is a control. A text block written for crawlers that no reader will ever see is the violation.

Pass 4: check every claim that can go stale

Prices, customer counts, product limits, dates, credentials, availability. Every hardcoded fact on a commercial page expires eventually, and a page that states an old price damages trust with the exact buyer your rankings are supposed to deliver.

List each changeable fact on the affected pages with its source and the date you last verified it. Fix what's stale. This pass rarely explains a rankings drop on its own, but it is the cheapest trust repair in the whole audit, and stale claims cluster on exactly the kind of neglected page families that spam systems flag for other reasons.

Pass 5: measure page families before you delete anything

Most post-update audits go wrong here, in both directions.

Google's scaled content abuse policy targets "many pages generated for the primary purpose of manipulating search rankings and not helping users", however they were created, AI or human. The doorway policy targets near-identical pages funnelling users to one destination. Neither policy says templates are spam. A shared frame across service pages, location pages or comparison pages is engineering, and buyers rely on it to find the same answers in the same place.

So measure the bodies, not the template. For each page in the family, count what is genuinely its own: distinct facts, local details, real examples, proof, page-specific internal links, a clear job. Read the weakest pages in full. A family where every page carries its own facts and serves a real query is coverage. A family where the pages swap a suburb name into identical paragraphs is the policy violation, no matter how nice the template is.

Vendor checklists that circulated after the August update ship exact thresholds for this: sentence-overlap percentages, pages-added-per-week limits, click floors for pruning. Google publishes none of those numbers. Use them as prompts for your own measurement if you like, but never let a vendor heuristic make a deletion decision a body-level read should make. This measured page-family discipline is the same standard we build to on every Agentic Website: big useful coverage, where every page earns its route.

One reversible change, then measure

The audit produces findings. Resist the urge to act on all of them at once.

Rank them by commercial risk, protect your money pages from broad experiments, and make one reversible change first: repair one page family, remove one genuinely dead section, fix one hidden-text pattern. Verify the change on the live site. Then measure the same page cohort against your frozen baseline over the following weeks.

Set honest expectations on the timeline. Google says improvement comes when its systems learn "over a period of months" that the site complies, and it told Schwartz that recovery from this update can take many months, with periodic refreshes along the way. Recovery is stepped, not linear, and a quiet month does not mean the fix failed.

One exception to plan around: link spam. Google is explicit that when its systems neutralise spammy links, any ranking benefit those links gave you is gone for good, and cleanup won't bring it back. The August update didn't target links, but if your history includes bought links, the recovery maths changes: you're not restoring old rankings, you're earning new ones.

AI Overviews are part of the same audit now

Google's spam policies now explicitly cover attempts to manipulate generative AI responses in Google Search. Content churned out to farm AI Overview citations is spam under the same policy as content churned out to farm blue links, and SEOs tracking the August rollout were watching exactly that wave of GEO and AEO spam.

Keep the observation separate, though. AI Overviews and AI Mode use different retrieval than the classic results, so record AI visibility on its own: the query, date, market, surface and which sources the answer cited. A lost AI citation and a lost ranking can arrive together without sharing a cause, and collapsing them into one number hides whichever problem you actually have. We keep those fields separate in every measurement we run, and the same rule is worth stealing at any size. Our guide to AI search metrics covers the full observation record.

How we run this at Searchmaxxed

Everything above is what we do for clients, minus the parts a public article can't do for you.

The normal agency response to a spam update is one of two failures. The panic agency bulk-deletes "thin content" in week one, destroys the evidence, and often deletes coverage the site needed. The report agency sends a PDF describing the update and changes nothing. Both leave the owner where they started, minus a month.

Our operating model is built so the audit is never a special event. The Managed Search Loop reads Search Console, analytics, live SERPs and AI-answer observations every week, ships one bounded improvement at a time through human approval, and verifies every change on the live site. When an update rolls out, the baseline already exists, the diffs are already recorded, and the question is just what changed this week. That loop is how a professional-services client of ours grew organic search clicks 719% across matched 90-day Search Console windows, with average position improving 10.9 places. The same evidence discipline that grows a site is the discipline that diagnoses one.

And the Off-Page Source Layer keeps the entity side aligned, so the facts Google and AI systems read about you away from your site corroborate what your pages claim. Sites with a clean, corroborated story are exactly the sites spam systems are built to reward. Case studies with the full numbers are at /case-studies.

FAQs

What is a Google spam update?

A spam update is a notable improvement to Google's automated spam-detection systems, including SpamBrain, its AI-based spam-prevention system. Google announces spam updates on its Search Status Dashboard and directs affected sites to review its spam policies.

Was there a Google update recently?

Yes. The August 2026 spam update ran from 18 August to 21 August 2026, globally and in all languages. It was the third spam update of 2026, following updates in March and June.

How do I know if the spam update hit my site?

Compare Search Console clicks, impressions and average position across the weeks before, during and after 18 to 21 August. A sustained drop that starts inside the window, on stable demand, is a candidate. There is no notification for algorithmic spam actions, so the data is the only signal.

How long does recovery from a spam update take?

Google says sites improve after its systems learn, over a period of months, that the site complies with its spam policies. It also confirmed periodic refreshes of the August update. Expect stepped progress across months, not a bounce the week after you fix things.

Should I file a reconsideration request?

Only for a manual action, which appears with a message in Search Console. Algorithmic spam updates have no reconsideration process. The fix is compliance plus time.

Is AI-generated content spam?

Not by itself. Google's scaled content abuse policy applies to mass-produced pages that exist to manipulate rankings rather than help users, "no matter how it's created". AI-assisted pages with real facts, original insight and a genuine job are not the target. Hundreds of near-identical pages are, whether a human or a model wrote them.

Are templated pages spam?

No. Templates are a design pattern, and consistent structure helps buyers. The policy question is whether each page in the family carries its own useful content. Measure the page bodies before deciding, and read the weakest pages in full.

Should I delete pages after a spam update?

Only after a page-level review supports it. Deleting in bulk on day one destroys the evidence you need for diagnosis, removes coverage and internal links you may still need, and can turn a recoverable dip into a real loss. Fix or remove page by page, with a recorded reason.

Is text inside accordions or tabs a violation?

No. Google's hidden text policy explicitly allows dynamic show-hide elements like accordions and tabbed content that serve readers. The violation is text placed solely for search engines that ordinary visitors can't reasonably see.

Did the August 2026 update target links?

Google said this update does not target link spam and does not target the site reputation abuse policy. Link spam is handled by separate systems, and Google notes that ranking benefit from neutralised spammy links cannot be regained.

What percentage of searches did the update affect?

Google declined to say. Any exact impact figure you see is a third-party estimate from that vendor's own tracking, not a Google number.

Can losing an AI Overview citation mean a spam problem?

It can accompany one, but it isn't proof on its own. AI Overviews use different retrieval than classic results, so record AI citations as a separate observation with the query, date, surface and cited sources, then diagnose each loss on its own evidence.

What to do next

Run the timing check today. It takes an hour in Search Console and it tells you whether you have an update problem, a demand problem or no problem.

If the data says your site was hit, or you'd rather know your exposure before the next update lands, our AI Visibility Audit covers this entire process across Google and AI answers: the evidence windows, the page-family measurement, the hidden-text and stale-claim passes, and a ranked, reversible fix plan. It's A$4,500, takes six business days, and the full amount is credited if you start a program with us within 14 days, so the audit effectively costs nothing if we do the repair work together. Book it here, or LET'S TALK if you'd rather show us the chart first.

Related reading: AI search metrics: what to measure · Case studies · The Managed Search Loop