Transcript
Intro
On the seventh of May 2026, FAQ rich results stopped appearing in Google Search. For every website on earth. That same week, agencies were still selling FAQ schema as a growth tactic. Some of them still are. Ask for a screenshot of the result.
That's modern technical SEO in a nutshell. Some of it still pays. Some of it pays nothing and keeps getting sold anyway. The trick is knowing which is which before you hand a developer a month of work.
So treat this episode as an effort budget for structured data, speed, international and discovery, ending with a one page cheat sheet of what earns what.
Structured data: what it earns
Structured data first, because it carries the most folklore. The claim you hear everywhere: schema markup is a ranking factor. Add it, move up.
Google answers this in a sneaky way. If your structured data is spammy, you can get a manual action, and Google spells out what that action does: the page loses eligibility for rich results, and it doesn't affect how the page ranks in web search. If even the penalty for bad schema leaves rankings alone, the markup was never a ranking lever. Verdict: placebo.
So what does it do? It makes a page eligible for specific rich results, like review stars, product prices or a video thumbnail. Google also says it uses structured data to understand what's on the page. Eligible is the key word, because Google says correct markup is still no guarantee the rich result shows.
Let's check it on a real tool. This is ReelRadar, our fictional film and where to watch site. Its film page carries Movie markup, VideoObject markup for the trailer, and an FAQ block that somebody added in 2022 because a blog post told them to.
Two invalid items, and the FAQ block isn't even listed any more, because Google stopped showing FAQ results. That's the first lesson: markup that looks fine in your code can still fail the test, and invalid markup earns nothing. The tool lists every missing or wrong field, so fix those first. And the FAQ block? It doesn't hurt. Google said back in 2023 that unused structured data causes no problems. It just sits there, like a gym membership.
The rich results Google killed
FAQ isn't a one off. Since 2023, Google has been trimming the list of rich results, step by step. Here's the timeline, all from Google's own announcements.
Notice the sentence Google repeats almost every time. When the sitelinks search box went, Google said it doesn't affect rankings. When seven types went in June 2025, it said the update won't affect how pages are ranked. If removing the rich result never moved rankings, adding the markup wasn't moving them either.
Second claim. FAQ schema gets you more space on the results page. This one was true for a few years, and plenty of sites got real clicks out of it. The only question today is whether the feature still exists.
It doesn't. Limited to government and health sites in August 2023, gone for everyone in May 2026, documentation deleted in June. If an audit or a plugin still tells you to add FAQ markup for Google, it's reading an old map. Keep the FAQ section if it helps readers. Just don't expect a rich result. Verdict: placebo.
And the AI angle, because you'll hear it next. Google's own guide to generative AI search lists overfocusing on structured data as a myth. In its words, there's no special schema markup you need to add. In the same breath, it says keep using structured data for rich result eligibility. Both halves matter.
So where is Google still investing? This year it added a creator property for videos in September, and product category and sale dates in July. Also in July, a review snippet rule against fake and undisclosed incentivized reviews. Commerce and video keep growing. Question and answer gimmicks keep shrinking.
That review rule earns its own warning. Review markup is where structured data most often gets sites a manual action, because people mark up ratings no real user ever gave. Google says ratings not by actual users may lead to a manual action. You lose the stars, and they only come back after a cleanup and a review request.
So the rule for structured data is short. Check Google's search gallery for features your page type can actually win. Mark those up accurately, and match the markup to what's visible on the page. Everything else is housekeeping, not growth.
How do you know it paid? Search Console has reports for the rich results it supports, and the Performance report can filter by search appearance. Compare click rate on pages with the rich result against similar pages without. If the feature never shows up, the markup is costing maintenance and returning nothing.
Core Web Vitals, right-sized
Now speed. Here Google says something stronger than it ever says about schema. Its page experience documentation states that Core Web Vitals are used by its ranking systems. So this one is a real input. The interesting question is how big.
Same page, same answer. Google always seeks to show the most relevant content, even if the page experience is sub-par. But when lots of helpful pages compete for one query, a great experience can contribute. Read that as an order of operations. Relevance first. Speed as a tiebreaker between pages that are close.
The thresholds, for the record. Largest Contentful Paint within two point five seconds. Interaction to Next Paint under two hundred milliseconds, which replaced First Input Delay in March 2024. Cumulative Layout Shift under zero point one. All judged at the seventy fifth percentile of real visits, mobile and desktop separately.
How are real sites doing? The Web Almanac's July 2025 data says forty eight percent of mobile sites pass all three. Now look at which metric holds them back. Responsiveness is good on seventy seven percent, layout shift on eighty one, and loading is the weakest at sixty two. On mobile, the usual villain is still a slow main image.
Where do you start? Not page by page. Search Console's Core Web Vitals report groups similar URLs together, so a failing group usually means a failing template. Fix the template that carries your traffic, the film pages on ReelRadar, and hundreds of URLs move at once. Then be patient, because field data covers the trailing twenty eight days.
If your problem is responsiveness, look at what runs when someone taps. The Web Almanac notes that deeper pages tend to carry filters, carousels and form checks, plus third party widgets and analytics that pile up JavaScript. On a film site, that's the trailer player and the ad stack. Remove or delay whatever blocks the main thread first.
Third claim, and the expensive one. You need a Lighthouse score of one hundred. We've seen teams spend a whole quarter chasing that green circle.
Here's the catch. Lighthouse is lab data: one simulated page load, on one device, on one fixed network. Search Console reports field data: real Chrome users over the last twenty eight days. Google's own PageSpeed docs say the two can disagree, and that good lab data doesn't mean real users are having a good time.
Here it is on a live page. PageSpeed Insights shows both in the same report, field at the top, lab further down. Watch how they can tell two different stories about one URL. When they disagree, the field data is what your users actually lived through, and it's the one Search Console reports.
Google even says it outright: chasing a perfect score just for SEO reasons may not be the best use of your time. A lab score of ninety and above already counts as good. Past that, the work stops paying in search. Verdict: placebo.
So the honest verdict on Core Web Vitals is works, modestly. Pass the field thresholds on the templates that earn traffic, then stop. It won't rescue a page that answers the wrong question. And the other page experience items, like HTTPS or avoiding pop ups, Google says don't directly help rankings. Do them for users.
hreflang: the rules that break silently
International next. hreflang tells Google which language or country version of a page to show to which searcher. What makes it dangerous is that it fails silently. Search Console's international targeting report was retired in 2022. Broken annotations don't throw errors. They just get ignored.
ReelRadar is a good example, because where to watch answers change by country. A film can be on one service in the United Kingdom and a different one in the United States. So ReelRadar runs a UK edition and a US edition of each film page. That's exactly the job hreflang exists for.
Rule one: return links. Each version lists itself and every other version. Google is blunt about it: if two pages don't both point to each other, the tags will be ignored. That's a safety feature. Otherwise anyone could declare themselves an alternate version of your page.
Rule two: valid codes. A language code, optionally followed by a country code, never a country on its own. The classic mistake is the United Kingdom, whose code is G B, not U K. Google's docs name this case directly: U K in hreflang has no effect on Google Search. We still find it on live sites every year.
Rule three: hreflang doesn't decide the language. Google says it doesn't use hreflang or the lang attribute to detect a page's language. It works that out itself. So a page tagged German that's mostly English is an English page. And add an x-default for everyone you didn't target.
You can declare hreflang in three places: link tags in the page head, an HTTP header, or the XML sitemap. Google treats them as equivalent. On big sites, the sitemap is often easier to keep complete, because one file holds every pairing. For PDFs, the header is your only option. Pick one method and keep it complete.
Two myths travel with hreflang. First, that it protects you from a duplicate content penalty. Translated pages were never duplicates. Google says localized versions only count as duplicates if the main content stays untranslated, so a German page with an English body is the real problem, and no tag fixes that.
Second, that machine translation is banned. It isn't. But mass translation that adds little value sits under scaled content abuse, which is next episode's territory.
Rule four, the one that quietly sinks international sites: redirecting visitors by IP address or browser language. Google says Googlebot mostly crawls from addresses that appear to be in the United States and sends no language header. Auto redirect, and Googlebot may never meet your other versions. Google recommends separate URLs with hreflang, plus a visible switcher. Verdict on auto redirects: placebo, with a side of harm.
Notice what hreflang is not. It's not a ranking boost, and it doesn't share authority between versions. It decides which version of a page shows to which searcher. If your UK visitors land on US prices, that's worth fixing, and that's what it fixes.
Here's what it looks like when all of this gets fixed at once. A dental clinic in Istanbul serving patients from abroad came to us running eleven languages of machine translation on query string URLs, plus a lot of spammy backlinks. We replaced the machine translation with proper translations, moved every language to clean URLs with a 301 from every old address, added hreflang across the set and cleaned up the link profile.
The receipts, from Search Console. Before our work, the clinic averaged about twenty clicks a day. On the twenty seventh of May, before we started, it got eight. On the seventeenth of September, it got fifty eight. September 2026 averaged thirty four point eight a day, up seventy seven percent, and daily impressions were up a hundred and twenty three percent. That's the whole programme at work, and the language fix is the foundation it stands on.
Sitemaps, IndexNow and Bing
Discovery last, and quickest, because most of the myths here are about tags Google ignores. Google's sitemap docs put it in one line: Google ignores priority and change frequency values. Every sitemap generator still writes them, and every audit still counts them. Verdict: placebo.
The tag Google does use is lastmod, on one condition: it has to be consistently and verifiably accurate. So a sitemap that stamps today's date on every page every night teaches Google to stop trusting your dates. Change lastmod when the content really changes. Only then.
And keep sitemaps in proportion. Google calls submitting one merely a hint, with no guarantee of crawling or indexing. A well linked site of about five hundred pages or fewer may not need one at all. Big sites split them at fifty thousand URLs. List canonical pages only.
One trick from our own client work that costs nothing: split your sitemap by page type. ReelRadar would have one for films, one for people, one for news. Search Console then shows indexing per sitemap, so when one page type stops getting indexed, you see which one the same day instead of digging through one giant list.
Now IndexNow. It's a protocol where you ping search engines the moment a page is added, changed or deleted, and one ping is shared with every participant. As of October 2026 the list is Bing, Yandex, Seznam, Naver, Yep, the Internet Archive and Amazon. Google is not on it.
Why bother with Bing at about five percent of search? Because Bing feeds more than its own results page. DuckDuckGo says it largely sources its traditional links from Bing. Microsoft Copilot answers with it. And since February 2026, Bing Webmaster Tools has an AI Performance report that counts how often Copilot and Bing's AI summaries cite your pages.
So here's the cheap win. Verify the site in Bing Webmaster Tools, submit the sitemap there, and switch on IndexNow if your platform supports it. One afternoon, every search product built on Bing, and zero cost on the Google side.
One correction while we're here. IndexNow doesn't speed up Google, because Google isn't a participant. Pinging it does nothing on that side. For Google, the honest levers are still internal links, a clean sitemap and pages worth crawling. Verdict: placebo.
The effort budget
Let's turn the whole episode into one budget. If you get one developer sprint for technical SEO this quarter, here's roughly where we'd spend it, in order. That's our judgement from client work, not a Google ranking.
And the order bends to the site. If you sell in several countries, a broken hreflang setup can cost more than everything else on this list combined, because the wrong version shows to the wrong market. Budgets follow the site, not the episode.
Wrap up
Here's the episode on one screen. The pattern holds all the way through: the things that change how a result looks, or which version shows, are real. The things sold as hidden ranking boosts mostly aren't.
The structured data cheat sheet, with every feature and what it can earn, is free on the episode page at irongoo.com. We update it whenever Google adds or removes a feature, which, as you saw, keeps us busy.
