01The March 2026 Core Update: What Really Happened

If you work in SEO and your traffic fell off a cliff in late March, you weren’t imagining things. The March 2026 Core Update — which ran March 27 through April 8 — was, by measurable data, the most volatile Google core update in recent history. Not by a small margin.

For context: the December 2025 Core Update — itself considered volatile — saw 66.8% of top-3 results shift. March 2026 hit 79.5%. That’s not a mild recalibration; that’s a significant raise of the bar.

Who won and who got buried

The clearest pattern from Amsive’s analysis of SISTRIX visibility data: Google moved traffic past the layer of sites that index, list, or comment on third-party inventory. Aggregators, comparison middlemen, parasite-SEO hosts — these took the worst hits across every vertical.

YouTube’s drop was remarkable enough to get its own paragraph. A -567 point SISTRIX visibility loss in under two weeks. That’s roughly 30% larger than Wikipedia’s historically unprecedented -435 point drop in December 2025. The exact mechanism isn’t confirmed, but the signal is hard to miss: even dominant UGC platforms aren’t immune to re-evaluation.

The update timeline you need to know

The spam update completing one day before the core update started created an unusual overlap. Some traffic changes from the spam update got falsely attributed to the core update, and vice versa. If you’re auditing your own losses, keep that timeline in mind — they’re separate systems doing separate things.

02Ranking Factors: What We Actually Know (and Don’t)

Here’s the uncomfortable truth I keep coming back to: nobody knows the exact weighting of Google’s algorithm. Not the biggest agencies, not Google’s own junior engineers (it’s compartmentalized), and definitely not the person selling you a “guaranteed #1 rankings” package on LinkedIn.

What we have are correlation studies, leaked documents, SERP analysis tools, and case studies from people willing to share their own data. I’ll show you three perspectives and then tell you what I actually believe.

What the major data sources tell us

What I actually believe, from running four content sites

My sample size is small. This isn’t science. But after eight years: updating old posts consistently drove 15–40% traffic increases within two months on three of my four sites. I genuinely don’t care whether that’s the freshness algorithm reacting or whether I’m just making the content better each time. It works. Do it.

The “freshness = 6% of ranking factor” number you sometimes see floating around? That’s one firm’s estimate, applied to their client sample. It could be 2%. It could be 15%. It almost certainly varies by niche and query type. Don’t quote it as a universal truth.

03Core Web Vitals: The Parts That Actually Matter in 2026

Google changed one of the three Core Web Vitals on March 12, 2024 — they replaced FID (First Input Delay) with INP (Interaction to Next Paint). If you haven’t audited your INP score since then, now’s the time. INP is the most-failed metric among the three.

Critical thing most guides gloss over: Google measures at the 75th percentile of real users. Not your lab tests in PageSpeed Insights. Not your fast broadband connection in a city. The actual 75th percentile of people visiting your site, on their devices, on their connections. A site that looks fine in testing can still fail CWV if a meaningful chunk of real visitors have a worse experience.

Real business impact — the numbers that actually matter

These case studies are from Google’s own documentation. The underlying principles still hold in 2026 even if the numbers shift:

How to fix the most common issues (copy-paste ready)

LCP — Loading performance:

<!-- Preload your hero image — this is often the LCP element -->
<link rel="preload" as="image" href="hero.webp">

<!-- Use WebP (25-35% smaller than JPEG/PNG) -->
<!-- Defer non-critical JavaScript -->
<script src="non-critical.js" defer></script>

INP — The one most people are failing: INP is about JavaScript blocking the main thread. Every click, tap, or interaction during the entire session gets measured — not just the first one like FID used to. The most common fix:

// Break up long-running JavaScript tasks
async function processLargeDataset(items) {
  for (let i = 0; i < items.length; i++) {
    processItem(items[i]);
    // Yield to the main thread every 50 items
    if (i % 50 === 0) {
      await new Promise(r => setTimeout(r, 0));
    }
  }
}

// Debounce expensive event handlers
window.addEventListener('resize', debounce(handleResize, 250));

Check the Coverage tab in Chrome DevTools. If you have large chunks of JavaScript that aren’t being used, remove them. This is the fastest INP win for most sites.

CLS — Stop your content jumping:

<!-- Always set explicit dimensions on images -->
<img src="photo.jpg" width="800" height="600" alt="description">

/* Reserve space for ads before they load */
.ad-slot { min-height: 250px; }

/* Use font-display: swap to prevent layout shift from web fonts */
@font-face {
  font-family: 'YourFont';
  src: url('font.woff2') format('woff2');
  font-display: swap;
}

04AI Search in 2026: Honest Data, No Hype

The “AI is killing SEO” discourse has been loud since 2023. Here’s what the actual data shows as of Q2 2026 — with the appropriate caveats attached to every number.

The real size of AI search right now

That conversion rate gap is the most important number in AI search right now. ChatGPT referral traffic is tiny in volume — roughly 4–7% of total web referrals for content-heavy sites — but it arrives with dramatically higher intent. Users have already been through a research/synthesis phase before clicking. They come pre-qualified.

Where AI referral traffic actually comes from

Not all AI platforms are equal for generating referral clicks. The breakdown as of early 2026, per multiple sources:

What I’m actually doing about AI search (and what I’m not)

I’m not rebuilding my entire content strategy around AI citations. Not yet. What I am doing:

Adding clear 40–60 word direct-answer paragraphs at the top of key pages. AI models pull from these. Adding FAQ sections in proper schema markup. Keeping my traditional SEO strong, because 76.1% of AI Overview citations come from pages already in the top 10 on traditional Google — so strong SEO remains the foundation.

What I’m not doing: chasing every “40,000% increase in AI mentions” case study from agencies selling AI visibility tools. Those numbers almost always measure mentions, not clicks, and usually compare against a near-zero baseline. Four thousand percent of almost nothing is still almost nothing.

05E-E-A-T: What Google Is Actually Rewarding Now

E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness — has been in Google’s Quality Rater Guidelines for years. What changed in 2025–2026 is that Google has moved from using it as a human evaluator framework to building algorithmic signals around it. The January 2026 algorithm update explicitly targeted AI-generated content without meaningful human contribution. The March 2026 update continued that pattern.

The hierarchy matters. Trust is the foundation. You can have all the expertise in the world, but if your site isn’t trustworthy — broken contact info, no HTTPS, no authorship transparency, claims that don’t hold up — none of the rest counts.

How to actually signal each component

Health, Finance, and Legal content (YMYL) — extra scrutiny

“Your Money or Your Life” content faces a higher bar because the stakes of bad information are higher. The difference between getting this right and wrong:

The difference is specificity: named journal, named authors, specific study design, exact sample size, precise numbers, and acknowledged limitations. That’s what Google’s systems are looking for in YMYL content, and it’s exactly what a person with genuine expertise naturally provides.

06Backlinks: What Still Works in 2026

Backlinks still matter for ranking in competitive niches. But “link building” as an industry is mostly dead. The game has changed: the sites winning today are being worth linking to, not buying or trading links.

HARO/Featured.com — The only mass-scale outreach that still works

Journalists send queries. You respond with expert insight. Success rate is 5–10% of responses, but when they land: links from major outlets that move rankings on their own. The template that actually works:

Subject: RE: [Query Topic] - [Your Specific Expertise]

Hi [Reporter's name],

I'm [Name], [Title] at [Company]. I've [relevant experience in 1 sentence].

On [their specific question]:

[2–3 sentences with a direct, specific answer including data]

Three specific points:
1. [Actionable insight with a concrete number or example]
2. [Counter-intuitive finding — something others wouldn't say]
3. [One thing most coverage gets wrong]

Available for follow-up questions.
[Name, title, phone]

Original research — the link magnet that still works

Surveys, data analysis, A/B test results. Not “5 tips for X” — original findings nobody else has. One study with a surprising conclusion and a clean methodology earns 10–100 quality links if it’s genuinely useful. Worth far more than any outreach campaign.

Broken link building — underrated and underused

Find relevant resource pages. Check for dead links (Screaming Frog or Ahrefs do this automatically). Create the resource the dead link was pointing to. Reach out. Success rate: 10–15% with genuinely good content as the replacement. The email that works:

Subject: Broken link on [Page Title]

Hi [Name],

I was using your [Topic] resource page at [URL].

One link is returning a 404: [broken URL]

I published a guide on this topic that covers [key points]: [your URL]

Thought you'd want to know.

[Name]

No pitch. No request. Just useful information. The link request is implicit — if they fix the dead link, they’ll often use yours.

07Schema & AI Optimization: 10-Minute Wins

Structured data does two things in 2026: it helps Google understand your content, and it helps AI models extract it. The latter is increasingly valuable. Here are the three schemas that give the best return for the time invested:

FAQ Schema — The AI model’s favorite format

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What is LCP in Core Web Vitals?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "LCP (Largest Contentful Paint) measures loading speed. Good score: under 2.5 seconds, measured at the 75th percentile of real users."
    }
  }]
}
</script>

Article Schema — The trust baseline

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Your Article Title",
  "author": {
    "@type": "Person",
    "name": "Your Name",
    "url": "https://yoursite.com/about"
  },
  "datePublished": "2026-05-09",
  "dateModified": "2026-05-09",
  "publisher": {
    "@type": "Organization",
    "name": "Your Site Name"
  }
}
</script>

Direct answer format — Structure that AI pulls from

## What is [Topic]?

[40–60 word direct answer. One clear statement.
No hedging. No "it depends." Lead with the core fact.]

### Why This Matters

[Detailed explanation with specifics, data, context.]

AI models almost always pull from the first direct answer they find on a page. If your opening paragraph is hedged, context-heavy preamble — “In today’s fast-moving world of digital marketing…” — you’ve already lost the AI citation race. Answer first. Expand second.

08My Personal Priority List for 2026

Eight years of running content sites comes down to this. Not a 50-point checklist. Not a comprehensive strategy deck. Just the things that actually moved the needle, ranked by what I’d do first if I had two hours this week:

09Sources — Verify Everything Yourself

Every number and claim in this post has a source. Here they are with honest notes about what each one is and what it isn’t.

No affiliate links. No sponsored content. No agency pitches. Just what the data shows, with honest notes on where the data is weak. If something here is wrong, let us know and we’ll correct it.