Two Optimization Tactics Everyone's Still Doing (That Google Killed in 2026)

We're three weeks into Q3 2026 and I'm still seeing proposals that recommend adding llms.txt to boost Google AI Overviews ranking. I'm still seeing agencies pitch FAQPage schema as a "quick win" for rich results. Both tactics are dead. Here's the breakdown.

llms.txt Doesn't Touch Google Search or AI Overviews

On June 15, 2026, Google clarified what should have been obvious: llms.txt has no effect on Google Search, AI Overviews, or AI Mode. Google's AI features use the same crawl index as ordinary Search — not a separate llms.txt feed.

The file was designed for platforms like Perplexity and ChatGPT Browse — systems that respect explicit AI-crawl directives. Google doesn't. If you want to control what Google's AI features see, you use the same levers you've always used: robots.txt, meta robots tags, structured data, and content quality signals.

Does llms.txt still have value?Yes — for non-Google platforms. Perplexity respects it. ChatGPT Browse respects it. If you're optimizing for those systems, llms.txt is a useful signal. But if your agency sold you on llms.txt as a Google AI Overviews boost, you got sold vapor.

Google Retired FAQ Rich Results — the Schema Isn't Dead

Google stopped showing FAQ rich results on May 7, 2026. The Search Console FAQ report was removed. The Rich Results Test dropped FAQPage support. API support ends in August 2026. It's not coming back.

Why did Google kill it?Abuse. Low-quality sites were stuffing FAQPage markup onto every page to grab SERP real estate. Google pulled the plug rather than police it. The deprecation was announced in August 2023 and rolled out over three years — this wasn't a surprise.

Is FAQPage dead?Not the markup — just the Google rich result. FAQPage is still the cleanest machine-readable way to mark a block of text as a question and its answer, and non-Google AI engines still read it to figure out what to cite when they don't have a Google-scale index to fall back on. If you already have it implemented, leave it in place — Google doesn't require removing markup just because it retired the rich result. Just stop adding FAQPage to chase a Google SERP feature that isn't coming back — that's the part your agency has wrong. And if they're still pitching FAQPage as a Google rich-result play in 2026, find a new agency.

What Actually Works for AI Visibility in 2026

Schema types that are active and useful for Google in mid-2026:

• Article — bylines, publish dates, publisher metadata. Essential for news and editorial content.
• Product — pricing, availability, reviews. Powers Google Shopping and product carousels.
• LocalBusiness — address, hours, geo coordinates. Shows up in Maps and local packs.
• Organization — brand identity, social profiles, logo. Builds entity recognition.
• BreadcrumbList — site hierarchy. Helps Google understand your information architecture.

Notice what's not on that list: HowTo and SitelinksSearchBox. Both are in the same bucket as FAQPage — Google retired the rich-result feature (HowTo in September 2023, the sitelinks searchbox in November 2024) but never asked anyone to remove the markup, and it still doesn't. Same rule applies to all three: don't add them chasing a Google SERP feature that's gone, don't rip them out if they're already there, and don't assume "Google killed the rich result" means "the schema is worthless" — non-Google AI engines still read structured markup even when Google's SERP has stopped rendering it.

Beyond schema, the signals that actually drive AI citation in 2026 are the same ones that drove traditional SEO: E-E-A-T content (experience, expertise, authoritativeness, trustworthiness), clean technical infrastructure (fast load times, mobile optimization, proper indexing), and brand authority (backlinks, mentions, social proof).

Adobe launched "Adobe Brand Visibility" at Cannes on June 17, 2026 — a 300M prompt dataset for measuring brand presence in AI outputs. Generative Engine Optimization (GEO) and AI Engine Optimization (AEO) are maturing from buzzwords into measurable disciplines. The tactics that work are the ones grounded in data, not the ones recycled from 2023 listicles.

Stop Optimizing for Ghosts

If your optimization checklist still includes llms.txt for Google or FAQPage schema for rich results, you're optimizing for features that don't exist. The AI search landscape is moving fast — what worked in 2023 is often irrelevant by mid-2026. Track the deprecations. Follow the platform announcements. Test what actually moves your visibility metrics.

And if you're hiring an agency, ask them when Google retired FAQ rich results. If they don't know, walk.

Sources

• TryChad.ai — Rejected finding: an AI-readiness audit recommended adding Accept to the Varyheader for markdown negotiation, citing acceptmarkdown.com — not a specification, just one author's documentation of a convention, and the fix is a no-op on a site that serves no markdown-negotiated responses, August 2026. Primary source. The same “optimizing for a ghost” pattern this article documents for llms.txt and FAQPage.
• Google Search Central — changes to HowTo and FAQ rich results, August 2023 (HowTo rich results ended; the FAQ removal timeline that completed in 2026 was announced here). developers.google.com
• Google Search Central — “Farewell, Sitelinks Search Box,” removal starting November 21, 2024; markup removal not required. developers.google.com