SEO in 2026
February 13, 2026 · 9 min read
I read a lot of SEO content. Most of it is garbage. This post is my attempt to write the thing I actually want to read: specific, honest, and free of “guaranteed #1 ranking” bullshit.
If you want the TL;DR: Google just changed how Discover works (separately from regular search), freshness matters more than it used to, and AI tools are becoming a real traffic source—but not as fast as the hype suggests. Everything else? Pretty much what’s been working for the last couple of years, just more refined.
Let’s get into it.
Google Split Discover from Regular Search (And Nobody’s Talking About It)
On February 5, 2026, Google did something weird. They released a core update specifically for Discover, not for regular search. First time they’ve ever done that.
Source: Google Search Central Blog – “An update to Discover” (Feb 5, 2026)
Here’s what changed:
1. Geographic matching got strict
U.S. users now see mostly U.S. publishers in Discover. Same pattern rolling out in other regions. If you’re a UK site trying to show up in Discover for U.S. users, you’re probably out of luck now.
2. Clickbait gets buried
Sensational headlines that don’t match the actual content? Google is actively demoting them. They specifically called out “misleading thumbnails” in the announcement.
3. Site-wide E-E-A-T matters more than individual posts
This is the big one. Over on Search Engine Roundtable, publishers were reporting 60-90% Discover traffic drops overnight. But their traditional search rankings? Completely fine.
The pattern: sites with a few viral posts but weak overall authority got hammered. Sites with consistent topic expertise kept their Discover traffic.
What this actually means:
If you’re chasing Discover traffic, you can’t just publish one banger article anymore. You need to prove you’re an authority on the topic across your entire site.
For most of us? Honestly, just focus on regular search. Discover was always unpredictable anyway.
Ranking Factors: What We Actually Know (And Don’t)
Here’s the uncomfortable truth: nobody knows exactly how Google’s algorithm works. Not me, not the “biggest” SEO firms, not even most Google employees (it’s compartmentalized).
What we have are correlation studies, leaked documents, and educated guesses. I’m going to show you three different perspectives, then tell you what I actually believe.
Perspective 1: First Page Sage (U.S. focus)
First Page Sage publishes annual correlation studies. Their 2025 report (latest available as of February 2026) analyzed thousands of sites.
Caveats: Their methodology isn’t public. Their sample is presumably their own clients. They claim to be “the biggest SEO firm in the U.S.” (their words, not verified).
Their findings:
- Publishing good content consistently: high correlation with rankings
- Keyword in title: moderate correlation, declining
- Backlinks: moderate correlation, declining over time
- User engagement signals: growing correlation
- Freshness (updates): jumped significantly in their 2025 data
- Link diversity: small but present correlation
Perspective 2: Semrush Sensor (Global data)
Semrush tracks ranking volatility across millions of keywords. Their 2024 data shows:
- Content relevance factors: consistently high impact
- Technical factors (speed, mobile): baseline requirements, not differentiators
- Backlink profile changes: slower correlation with rankings than 5 years ago
Perspective 3: My Own Sites (Sample size: 4, bias: high)
I manage four content sites. Here’s what actually moved the needle for me in 2024-2025:
- Updating old posts: +15-40% traffic within 2 months (3 out of 4 sites)
- Fixing Core Web Vitals: zero noticeable impact (but I sleep better)
- Adding author bios: no measurable change (but seems right)
- Getting one BBC link: +200% traffic to that specific page
Important: This is n=4. Don’t treat it as science. Treat it as “maybe worth testing on your site.”
What I Actually Believe
Freshness matters more than 2 years ago. How much more? No idea. But when I update posts, they climb. Correlation? Yes. Causation? Maybe Google likes fresh signals, maybe I’m just making the content better. I don’t care – it works.
Backlinks still matter for competitive terms. But “link building” as an industry is mostly dead. What works: being worth linking to. Original research, strong opinions, useful tools. Everything else is noise.
User engagement is either a ranking factor or a side effect of good rankings. Chicken and egg. Either way, write for humans.
The “6% freshness” number you might have seen? That’s one firm’s estimate. It could be 2%, it could be 12%, it could vary wildly by niche. Don’t quote it as fact.
Core Web Vitals: The Parts That Actually Matter
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).
Source: web.dev – “Optimize Interaction to Next Paint” (implemented March 2024)
Here’s what you need to hit:
| Metric | What It Measures | Good | Bad |
|---|---|---|---|
| LCP – Largest Contentful Paint | Loading speed | Under 2.5s | Over 4.0s |
| INP – Interaction to Next Paint | Responsiveness | Under 200ms | Over 500ms |
| CLS – Cumulative Layout Shift | Visual stability | Under 0.1 | Over 0.25 |
LCP = when your main content loads (usually hero image or main text block).
INP = how fast your page responds to ALL user interactions, not just the first one. Every click, every tap, throughout the entire session.
CLS = whether stuff jumps around while loading (ads pushing content down, etc.).
Important: Google measures at the 75th percentile of real users. Not lab tests. 75% of actual people visiting your site need to have a “good” experience.
Does This Actually Matter?
Yeah. Real companies with real results:
Yahoo Japan (79 billion monthly pageviews) fixed its CLS issues:
- +15.1% pageviews per session
- +13.3% session duration
- -1.72% bounce rate
Pinterest improved LCP from 2.5s to 1.5s:
- +15% SEO traffic
- +15% sign-ups
Vodafone improved LCP by 31%:
- +8% sales
- +15% cart conversion
Note: These case studies are from 2021. The principles still apply, but your mileage may vary in 2026.
How to Actually Fix This (Copy-Paste Ready)
For LCP (Loading):
<!-- Preload your main image -->
<link rel="preload" as="image" href="hero-image.webp">
<!-- Use WebP images (25-35% smaller than JPEG/PNG) -->
<!-- Defer JavaScript -->
<script src="whatever.js" defer></script>Get better hosting. Seriously. Shared hosting will always struggle here. Use a CDN (Cloudflare’s free tier works fine).
For INP (Responsiveness):
This is about JavaScript blocking the main thread.
// Break up long tasks
async function processStuff(items) {
for (let i = 0; i < items.length; i++) {
doThing(items[i]);
// Yield to main thread every 50 items
if (i % 50 === 0) {
await new Promise(r => setTimeout(r, 0));
}
}
}
// Debounce expensive operations
function debounce(func, wait) {
let timeout;
return (...args) => {
clearTimeout(timeout);
timeout = setTimeout(() => func(...args), wait);
};
}
window.addEventListener('resize', debounce(handleResize, 250));Remove unused JavaScript. Check the Coverage tab in Chrome DevTools.
For CLS (Layout Stability):
<!-- Set image dimensions -->
<img src="pic.jpg" width="800" height="600" alt="thing">/* Reserve space for ads before they load */
.ad-slot {
min-height: 250px;
}
/* Use font-display: swap for web fonts */
@font-face {
font-family: 'Whatever';
src: url('font.woff2') format('woff2');
font-display: swap;
}AI Search: Growing, But Not There Yet
Everyone’s talking about “AI traffic” in 2026. Here’s what that actually means, with the hype filtered out.
What “AI Traffic” Actually Is
Right now (February 2026), there are three ways people find you through AI:
- Brand mentions in AI responses. ChatGPT or Perplexity mentions your company when answering a question. This is visibility, not necessarily traffic.
- Referral clicks from AI platforms. Some AI tools now link to sources. Volume is tiny compared to Google Search—for now.
- Traditional search traffic that started with an AI query. Hard to track, probably overlaps with normal SEO.
The “for now” is important. Perplexity, ChatGPT Search, and Gemini are growing fast. In 6-12 months, this picture could look very different. Don’t ignore it, but don’t panic either.
The Case Studies You’ve Seen (Deconstructed)
Brand24/Base case: “48.6% increase in AI visibility”
What this actually means: Brand24 has a tool that counts how often Base appears in AI responses. Brand24 sells this tool. Base is their client. The methodology isn’t public. The “traffic” might just be mentions, not clicks.
Source: Brand24 blog (Jan 26, 2026) – company self-reporting using its own tool
Promodo/Techtoro case: “4000% increase in AI mentions”
What this actually means: Techtoro went from roughly zero AI mentions to some AI mentions. 4000% of almost zero is still almost zero in absolute terms. Promodo is an agency selling SEO services. Techtoro is their client.
Source: Promodo “SEO Trends 2026” report – agency case study, baseline not disclosed
I’m not saying these case studies are fake. I’m saying the numbers are meaningless without context, and the context is “companies selling services to improve these numbers.”
What I’m Actually Doing About AI Search
- Adding FAQ sections to key pages. AI models seem to pull from structured Q&A.
- Using clear, extractable definitions. The first paragraph answers the question directly.
- Keeping traditional SEO strong. AI models favor sites that already rank well in Google.
- Watching the space. In February 2026, Google sends 100x more traffic than all AI tools combined. But I’m checking my analytics monthly to see if that ratio shifts.
E-E-A-T: What Google Actually Wants (Without the Acronym Worship)
E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trustworthiness. Google’s Quality Raters use these guidelines to evaluate search results.
Source: Google Search Quality Rater Guidelines (PDF) (updated 2024)
Here’s the thing nobody tells you: Trust is the foundation. The other three build on it. Untrustworthy pages have low E-E-A-T, no matter how experienced or expert they seem.
How to Actually Show This
Experience – First-hand knowledge
Show you’ve actually done the thing:
- Use “I tested” not “users report”
- Include your own photos, not stock images
- Share specific details only an insider would know
- Mention mistakes you made
Expertise – Deep understanding
Show you know what you’re talking about:
- Explain WHY things work, not just WHAT to do
- Use technical terms correctly (but explain them)
- Cite recent research
- Acknowledge what’s debated or uncertain
Authoritativeness – Others recognize you
This is the hardest to build:
- Get mentioned in industry publications
- Earn backlinks from respected sites
- Speak at conferences
- Get cited by other experts
Trustworthiness – You’re reliable
The foundation:
- HTTPS everywhere
- Clear contact info
- Real author photos and bios
- Transparent about affiliates/sponsors
- Correct errors when you find them
- Actual privacy policy
For Health/Finance Content (YMYL)
“Your Money or Your Life” topics get extra scrutiny.
If you’re writing about health, finance, legal stuff, or safety:
- You NEED verified credentials (MD, CPA, JD) or a “reviewed by” expert
- Cite multiple authoritative sources for every claim
- Include disclaimers
- Update more frequently
Example of doing this right:
❌ Instead of: “Studies show Mediterranean diet reduces heart disease.”
✅ Do this: “A 2024 meta-analysis in the Journal of the American College of Cardiology (Smith et al.) examined 15 randomized controlled trials with 12,461 participants. Adherence to a Mediterranean diet was associated with a 23% reduction in cardiovascular events over 5.2 years. However, this was strongest in European populations and less pronounced in other groups.
Disclaimer: This is educational info. Talk to your doctor before changing your diet.“
See the difference? Specific source, study design, sample size, exact numbers, limitations acknowledged, and disclaimer included.
Backlinks: What Still Works in 2026
Backlinks still matter for ranking in competitive niches. But the game changed.
What Makes a Good Link Now
Tier 1 – Editorial links (the gold standard):
- News sites mentioning your research
- Industry publications citing your data
- Journalist quoting you in an article
Tier 2 – Resource links:
- University course syllabi
- Government resource pages
- Industry association directories
Tier 3 – Community links (minor value):
- Thoughtful comments on relevant posts
- Social profiles
Avoid entirely:
- Link exchanges
- Paid links without
rel="sponsored" - Private blog networks
- Comment spam
What Actually Works
1. HARO (Help a Reporter Out) / Featured.com
Journalists send out queries. You respond with expert insights.
Success rate: 5-10% of responses get published. But when they do: links from major outlets.
Template:
Subject: RE: [Query Topic] - [Your Expertise]
Hi [Reporter],
I'm [Name], [Title] at [Company]. I've [relevant experience].
On [specific question]:
[2-3 sentence direct answer with data]
Three specific insights:
1. [Actionable point]
2. [Specific example]
3. [Counter-intuitive finding]
Available for interview [days/times].
[Your Name]2. Create Actually Original Research
Surveys, data analysis, A/B tests.
Examples:
- “2026 [Industry] Benchmark: 500 Companies Analyzed”
- “[Topic] User Behavior: What 10,000 Sessions Revealed”
Can earn 10-100+ quality links if genuinely useful.
3. Broken Link Building
Find relevant resource pages. Check for dead links. Create better content. Reach out.
Email:
Subject: Broken link on [Page Title]
Hi [Name],
I was using your [Topic] resource at [URL].
One link returns a 404: [broken URL]
I published a guide on this topic: [your URL]
It covers [key points].
Thought you'd want to know.
[Name]Success rate: 10-15% with good content.
AI Optimization: Schema That Actually Helps
AI models extract structured data. Give them what they want.
FAQ Schema
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "What is LCP?",
"acceptedAnswer": {
"@type": "Answer",
"text": "LCP (Largest Contentful Paint) measures loading performance. Good: under 2.5 seconds."
}
}]
}
</script>HowTo Schema
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "How to Improve LCP",
"step": [{
"@type": "HowToStep",
"name": "Preload Critical Resources",
"text": "Add preload tags for LCP image in your HTML head."
}]
}
</script>Article Schema
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your Title",
"author": {
"@type": "Person",
"name": "Your Name"
},
"datePublished": "2026-02-13",
"dateModified": "2026-02-13"
}
</script>Direct Answer Format
Structure content for extraction:
## What is [Question]?
[40-60 word direct answer]
### Why This Matters
[Detailed explanation]AI models often pull the first direct answer.
What Actually Matters (My Personal Priority List)
After doing this for 8 years, here’s what I’m focusing on in 2026:
- Update old content quarterly
When I update posts, they climb. Don’t care if it’s “freshness algorithm” or “I made it better” – it works. - Answer the question in the first paragraph
People want the answer fast. Give it to them, then expand. - Show your work
Cite sources. Use specific numbers. Admit what you don’t know. - Make your site fast enough
Core Web Vitals matter for user experience. Good UX correlates with good rankings. Fix the obvious stuff, don’t obsess over 50ms. - Build real relationships
One link from a real publication beats 100 directory listings. - Add basic schema
Takes 10 minutes. Might help with AI extraction. Definitely doesn’t hurt. - Test on your own site
What works for me might not work for you. What works in my niche might fail in yours. The only truth is your own data.
Sources & How to Verify
| Source | What It Is | Link | Caveat |
|---|---|---|---|
| Google Search Central | Official Google announcements | https://developers.google.com/search/blog | Primary source |
| First Page Sage | SEO correlation study (2025) | https://firstpagesage.com/seo-blog/ | Methodology not public |
| Semrush Sensor | Ranking volatility data | https://www.semrush.com/sensor/ | Third-party tool |
| Google Quality Rater Guidelines | Official evaluator guide | https://guidelines.raterhub.com/searchqualityevaluatorguidelines.pdf | Primary source |
| web.dev Core Web Vitals | Metrics & thresholds | https://web.dev/vitals/ | Primary source (Google) |
| Brand24 case study | Base AI visibility | https://brand24.com/blog/ | Company self-reporting |
| Promodo report | Techtoro case study | https://promodo.com/ | Agency case study |
| Yahoo Japan case | Performance optimization | https://web.dev/case-studies/yahoo-japan-news/ | Dated 2021 |
| Pinterest case | LCP improvements | https://web.dev/case-studies/pinterest/ | Dated 2021 |
| Vodafone case | Speed optimization | https://web.dev/case-studies/vodafone/ | Dated 2021 |
What I Don’t Know
- Exact percentage weights of Google’s algorithm (impossible)
- Whether “AI traffic” will matter in 12 months
- If the February 5 Discover update affects all regions equally
- Whether my own experiences generalize to your site
No affiliate links. No sponsored content. Just what I’ve learned, with the honesty about what I haven’t.
