Ultimate Guide to SERP Features and Rich Results (2026)

The 2026 Google results page is a mixed bag of things you can earn and things you can only influence. Rich results such as recipes, products, events, and reviews are earned with structured data, then measured in Search Console enhancement reports. Featured snippets, AI Overviews, People Also Ask, and knowledge panels are the opposite: Google decides to show them, no schema opt-in exists, and the best you can do is optimize the page and wait. This guide draws a hard line between the two, dates every deprecation that retired a rich result type, and walks three worked examples through live tools so you stop chasing result types Google already killed and start winning the ones that are still winnable.

What Are SERP Features vs. Rich Results? The Controllable/Uncontrollable Split

A SERP feature is any element on a search results page that is not a plain blue link with a two-line description. A rich result is a narrower, better defined thing: an enhanced search result that Google enables only after reading structured data on the page. The distinction is the whole game, because one side is controllable and the other is not.

Structured data is a standardized format built on the schema.org vocabulary that tells Google what a page is about and classifies its content, and Google uses that classification to enable rich results and other display treatments. Google structured data documentation (2025) defines it that way, and the operative word is "enable": markup unlocks a result type, but Google still decides whether to show it. A rich result is therefore an opt-in. You add the JSON-LD, you validate it, and you become eligible. Whether the star rating, the cook time, or the event date actually renders is then a function of your page quality and Google's own thresholds.

A SERP feature such as a featured snippet, an AI Overview, or a People Also Ask box has no opt-in at all. Featured snippets are the clearest case. Google describes them as special boxes that reverse the regular result format, showing the descriptive snippet first, and states that Google selects them automatically from web search listings with no special markup or opt-in. Google featured snippets documentation (2025) could not be plainer: you cannot apply for a snippet. You can only rank, structure the page, and either accept or decline the display through snippet opt-out controls, which a later section covers in full.

The split matters because it changes where you spend time. Rich result work is deterministic: write markup, validate it, track eligibility. SERP feature work is probabilistic: improve relevance and authority, then measure what Google chooses to surface. Teams that treat both the same way waste months adding schema in the hope of winning a featured snippet, a result type that ignores schema entirely. The table below is the spine of this guide.

Read the last column as the entire thesis: rich results are winnable, SERP features are influenceable. Everything after this section is either a playbook for the winnable column or a signal list for the influenceable one.

The Complete 2026 SERP Feature Map: Feature, Opt-In Path, and Whether You Can Win It

Google does not publish a single official catalog of SERP features, and the count changes constantly as Google tests, ships, and retires layouts. Third-party roundups tally dozens; SE Ranking's running list puts the number at 37 or more distinct features. SE Ranking SERP features list (2026). The exact total is less useful than knowing which features respond to which input. The map below groups every major 2026 feature by its opt-in path, because the opt-in path tells you whether the feature is worth your time.

Features with a structured-data opt-in

These are the rich results: recipe, product, review snippet, event, course, job posting, video, breadcrumb, article, local business, organization, and software app. Each has a schema.org type you add, a Rich Results Test pass, and a Search Console enhancement report. Google Search Gallery (2025) is the canonical list of which types are documented and supported. These are the only features you can genuinely "claim," and even here the claim is eligibility, not a guarantee.

Features with a content and authority opt-in

Featured snippets and People Also Ask both pull from pages that already rank well. There is no form to file. The input is relevance, ranking position, and a page structure that answers one question cleanly. You optimize signals, then Google decides. The practical difference from rich results is that you can do everything right and still not get the snippet, which is why the measurement for these features is manual SERP checks, not a console report.

Features with a product or entity opt-in

The local pack, the knowledge panel, and shopping results draw from separate systems. The local pack reads from Google Business Profile, the knowledge panel from Google's entity graph and corroborating web sources, and shopping results from Merchant Center feeds. Schema can help corroborate, but the primary input is not markup on your page. These features belong in a local SEO or entity strategy, not a rich result playbook.

Features with no opt-in at all

AI Overviews and AI Mode sit here. Google states that to appear as a supporting link in AI Overviews or AI Mode, a page must simply be indexed and eligible to be shown in Google Search with a snippet, with no additional technical requirements. Google AI features documentation (2025). That single sentence kills most of the AEO industry: there is no AI opt-in, only the same indexing and snippet eligibility you already need for ordinary search.

Every Google Rich Result Type Still Available in 2026 (and the Ones That Are Gone)

Google's structured data documentation defines a distinct set of rich result types, and the 2026 list is shorter than most ranking guides pretend. The documented types are Article, Breadcrumb, Course, Dataset, Event, FAQ, JobPosting, Local Business, Organization, Product, Recipe, Review snippet, Software App, and Video. Google Search Gallery (2025). Three of those carry a major asterisk, and one, FAQ, is effectively dead for most sites. The table below splits the list into what is actually winnable, what is restricted, and what is gone.

The active types are a small, well-understood set

Notice how few of the types are broadly relevant. For most commercial sites the realistic targets are Article, Breadcrumb, Product, Review snippet, Recipe, Event, and Video. If the site is a SaaS product, the SaaS SEO guide maps those types onto a product-led content engine. The rest matter only if you publish courses, datasets, jobs, or software. The cleanest way to shortlist is to look at what your pages actually contain and match the type to the content, which Step 2 turns into a decision framework.

FAQ is restricted, not removed

FAQ markup still exists in the documentation, but eligibility collapsed in August 2023. Starting that month, Google shows FAQ rich results only for well-known, authoritative government and health websites, and no longer regularly for all other sites. Google FAQ and HowTo changes (2023). If you run an ecommerce store or a blog, FAQ rich results are not coming back, no matter how clean your markup. A dedicated section later walks through the full deprecation timeline.

HowTo is gone entirely

HowTo rich results are deprecated. Google first limited them to desktop in August 2023, then stopped showing them on desktop entirely as of September 13, 2023. Google FAQ and HowTo changes (2023). Any guide still telling you to add HowTo markup to win a rich result is recycling a 2022 playbook. The markup does not hurt, but it no longer produces a rich result for anyone.

Step 1: Benchmark Your Current SERP Footprint in Search Console and Manual SERP Checks

Before you add a single line of markup, measure what you already have. The benchmark has two halves: Search Console tells you what rich results you are eligible for, and manual SERP checks tell you which emergent features you already appear in. Both are free, and both take an afternoon.

Read your enhancement reports first

Open Search Console and find the enhancement reports under the Performance area. Each report corresponds to a rich result type: Recipes, Products, Review snippets, Events, Breadcrumbs, and so on. A report that lists your pages means Google has already parsed your markup and judged it eligible. A report that is empty means you either have no markup or it has not been validated. This is your ground truth, and it is the number the rest of this guide will move.

Run manual SERP checks for the emergent features

Search Console cannot tell you whether you own a featured snippet, a People Also Ask position, or an AI Overview citation, because none of those have a report. You have to look at the results page itself, ideally in a clean browser session without personalization. Search your target queries, note which features render, and record whether your page appears in each. A SERP preview tool helps you sanity-check how your own title, snippet, and structured data will render before you go live, but it does not replace the live check for whether Google chose you for a feature.

Record a baseline you can re-check

Write the baseline down: for each of your ten highest-value queries, note today's position, the rich result types on the page, and whether you appear in any feature. Without that record, a later "we won the snippet" is just memory. Re-run the same check monthly. The keyword research tool gives you the query set to track, and the manual check gives you the feature set to watch.

Step 2: Choose the Right Rich Result Type for Each Page (Decision Framework)

Most markup mistakes are mismatch mistakes: a Recipe type on a page that is really a product listing, a Product type on a blog post, or a FAQ page on a marketing article. The decision framework is a three-question filter that forces the page and the type to agree.

Question one: what does the page actually contain?

The type must describe the visible content, not the content you wish you had. Google requires that structured data describe the page content that a user actually sees, and markup that describes hidden or unrelated content is a violation that can trigger a manual action. Start from the page as it is: a full recipe with ingredients and steps, a product with a price and an offer, an event with a date and a venue. If the visible content is not there, add the content before you add the markup.

Question two: is the type eligible for your site?

Some types carry site-level restrictions. FAQ rich results now apply only to government and health sites. Review snippets cannot be self-serving, meaning a site cannot mark up reviews it wrote about itself. A Product type requires one of review, offers, or aggregateRating. Check the eligibility column in the Step 3 table before you commit, and skip types your site does not qualify for.

Question three: is this the highest-value type for the query?

One page can support multiple types, but you should lead with the one that matches the search intent. A recipe page that also carries breadcrumbs and article markup is fine; a recipe page that also tries to be a product is not. Choose the type that answers the query, then layer breadcrumbs and organization markup on top as supporting types. The schema checker will flag types that conflict or lack required properties, so run it early rather than after publishing.

Step 3: Write schema.org JSON-LD Markup for Your Page

JSON-LD is the format Google recommends over microdata and RDFa, and it is the only one you should ship in 2026. It lives in a script tag in the head or body, it does not alter your visible HTML, and it is easy to template. The two blocks below are the anchors of this guide: a Recipe block and an Article block, both reduced to the minimum properties that pass validation.

A minimal, valid Recipe block

The Recipe type requires name, image, recipeIngredient, and recipeInstructions. Everything else, from cook time to nutrition, is optional enrichment. Ship the four required properties first, and only add optional ones when the data is genuinely on the page. The block below is a complete, valid example.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Recipe",
  "name": "Sourdough Boule",
  "image": ["https://example.com/images/sourdough-boule.jpg"],
  "author": {"@type": "Person", "name": "Maya Chen"},
  "datePublished": "2026-08-01",
  "description": "A 24-hour naturally leavened boule with a crackling crust.",
  "prepTime": "PT30M",
  "cookTime": "PT45M",
  "totalTime": "PT24H",
  "recipeYield": "1 loaf",
  "recipeCategory": "Bread",
  "recipeCuisine": "American",
  "nutrition": {"@type": "NutritionInformation", "calories": "180 calories per slice"},
  "recipeIngredient": [
    "500g bread flour",
    "350g water",
    "100g active sourdough starter",
    "10g salt"
  ],
  "recipeInstructions": [
    {"@type": "HowToStep", "text": "Mix flour, water, and starter. Rest 30 minutes."},
    {"@type": "HowToStep", "text": "Add salt and knead. Bulk ferment for 4 hours."},
    {"@type": "HowToStep", "text": "Shape and proof overnight in the refrigerator."},
    {"@type": "HowToStep", "text": "Bake in a preheated Dutch oven at 240C for 45 minutes."}
  ]
}
</script>

A minimal, valid Article block

The Article type needs headline, author, datePublished, and image. The example below is what you would place on this very guide or any editorial page. Note that the description mirrors the meta description, and the dates are ISO 8601 strings.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Ultimate Guide to SERP Features and Rich Results (2026)",
  "author": {"@type": "Organization", "name": "SEO.to"},
  "datePublished": "2026-08-16",
  "dateModified": "2026-08-16",
  "image": ["https://seo.to/images/serp-features.jpg"],
  "description": "Every SERP feature that matters in 2026 and how to win each one."
}
</script>

Two rules govern both blocks. First, every property must be present in the visible page content: the recipe ingredients and steps must actually appear on the page, and the article headline must match the H1. Second, keep the markup machine-readable and clean, with no unescaped characters inside the JSON. When you are unsure whether your block is well-formed, paste it into the schema checker before you commit it to the template.

Step 4: Validate Your Markup in the Rich Results Test

Validation is not optional, and it is not the same as well-formed JSON. A JSON block can parse perfectly and still fail the Rich Results Test because it is missing a required property or applies the wrong type. The test checks both, and it is the gate you must pass before Google will even consider your page eligible.

Run the test on the rendered page, not the snippet

Google's Rich Results Test accepts either a URL or a code snippet. Test the live URL whenever the page is public, because the test fetches and renders the page the way Googlebot does. If your JSON-LD is injected by JavaScript, the snippet view will pass while the rendered page fails, and the rendered page is the one that counts. When you are still drafting, paste the snippet, then re-test the URL after publishing.

Read the two failure classes

The test reports two kinds of problems: errors, which block the rich result, and warnings, which degrade it but do not block it. A missing required property is an error; a recommended but absent optional field is a warning. Fix every error, then decide on warnings case by case. A warning on cook time is worth fixing on a recipe page and irrelevant on an article page. The schema checker gives you the same required-versus-optional breakdown inline, so you can fix errors before you open the Google test.

Validate before every major change

Treat validation as part of the deploy, not a one-time pass. Templates drift: someone removes the image field, a CMS strips the script tag, a migration drops the JSON-LD. Re-run the test after any change that touches the page template, the markup, or the CMS. It takes one minute and catches the regressions that silently kill a rich result.

Step 5: Ship, Then Track Eligibility in Search Console Enhancement Reports

Shipping the markup is the midpoint, not the finish. After you publish, Google has to recrawl the page, re-read the JSON-LD, and re-evaluate eligibility. That takes days, not minutes, and it is exactly what the enhancement reports are for.

Confirm the page is indexed with the new markup

First, confirm the page is indexed and that the rendered page still contains the script tag. Use URL Inspection to request indexing, then check the rendered HTML that Google fetched actually includes your JSON-LD. A common failure is a caching layer or a tag manager that serves the markup to your browser but not to Googlebot. The on-page SEO checker prints the live heading outline and can confirm the page structure Google is reading, which closes that gap.

Watch the enhancement report appear

Once Google has recrawled, the corresponding enhancement report in Search Console will list the page under valid or invalid. Valid means the page is eligible; invalid means it has errors, and the report will name them. The report also lists pages that are valid with warnings, which are eligible but degraded. Your job now is triage: fix invalid pages first, then warnings, then move on.

Eligibility is not a guarantee of display

Here is the part most guides skip: a valid page in an enhancement report is eligible, not guaranteed to render as a rich result. Google still applies its own quality thresholds and may choose to show a plain link for a page that passes every test. The enhancement report measures the markup, not the display. If the report says valid and the SERP shows no rich result, the next section on debugging tells you where to look.

Step 6: How to Win Featured Snippets (No Opt-In Exists — Optimize These Signals Instead)

Featured snippets are the most misunderstood feature in SEO, because so many guides imply you can capture one with markup. You cannot. Google selects snippets automatically from already-ranking web results, with no special markup or opt-in. Google featured snippets documentation (2025). The word "selects" is doing the work: the snippet is a consequence of ranking and structure, not a target you file for.

Rank first, then structure for extraction

Featured snippets are pulled from pages that already rank, almost always on page one. So the first signal is ordinary ranking: the page has to be the best answer to the query before it can be lifted into the box. Once it ranks, structure decides whether Google can extract a clean answer. The page should state the answer in a short, self-contained sentence near the top, then support it. For list, step, and comparison queries, a tight table or ordered list often becomes the snippet. Google extracts the snippet text from the page; you are writing the page so that extraction is trivial.

Match the snippet format to the query type

Snippets come in three dominant shapes: paragraph, list, and table. A "what is" or definition query favors a paragraph snippet. A "how to" query favors a list snippet. A "versus" or comparison query favors a table snippet. Match your page structure to the query, and you make the extraction easy; fight the format, and Google will either skip you or cut a mangled excerpt. The keyword research tool tells you which queries carry snippets, and the query wording tells you which format to build for.

A before and after rewrite for one query

Take the query "what is a featured snippet." The before version opens with background and buries the definition three paragraphs down:

Before: "Search results have changed a lot over the years, and one of the most visible changes is the special box at the top of the page that many SEOs talk about constantly."

The after version states the answer in the first sentence, in plain terms, so Google can lift it whole:

After: "A featured snippet is a special search result box that shows a descriptive snippet from a web page before the regular results, and Google selects it automatically with no markup or opt-in required."

The after version is not keyword-stuffed and not gimmicky. It is a single self-contained answer that happens to be exactly what Google's documentation describes, which is the whole technique: write the page so the answer is extractable. For more on the on-page mechanics behind that extraction, the on-page SEO guide covers heading structure and answer placement in detail.

Step 7: How to Earn AI Overviews Citations (Grounding, Not Gimmicks)

AI Overviews citations are the 2026 successor to the featured snippet obsession, and the same discipline applies: there is no opt-in, only ranking plus extractability. Google's generative AI features are rooted in its core Search ranking and quality systems, using retrieval-augmented generation, called RAG or grounding, and a query fan-out technique to surface prominent, clickable links to supporting pages. Google AI optimization guide (2025). Read that sentence carefully: the AI cites pages that the core ranking systems already consider authoritative, then grounds its answer in them.

The entry requirement is ordinary snippet eligibility

The bar for being cited is the same bar as being shown in search at all. To appear as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet, and there are no additional technical requirements. Google AI features documentation (2025). That means the work is the work you already do: get indexed, pass the snippet bar, and rank. There is no AI layer to optimize on top of it.

Grounding means a page that already ranks with a clean snippet

Because the AI grounds its answer in retrieved pages, the best citation candidate is a page that already ranks for the question and carries a snippet Google can lift. The recipe is the same as the featured snippet recipe, applied to the questions your audience asks. A page that ranks third for a question and states the answer in a clean first paragraph is a far better AI Overviews candidate than a page that ranks for nothing but carries a dozen gimmicks. The Measuring AI Visibility section later in this guide extends this into a full measurement plan, but the core is here: rank, be extractable, and stop looking for a special AI on-switch.

A worked citation case

Consider a page answering "how long does sourdough need to proof." The page already ranks in the top five for the query. The fix that earns the citation is not a new file or a new tag. It is a rewrite of the opening paragraph so that the answer, a range of hours with the temperature context, sits in the first two sentences in plain prose. After the rewrite, the page becomes the grounding source Google lifts, and the citation follows. The page ranked before and ranks after; what changed is that the answer is now extractable. That, in two sentences, is the entire discipline of AI Overviews SEO.

Snippet Opt-Out Controls Explained: nosnippet vs. data-nosnippet vs. max-snippet

Google offers snippet controls for site owners who want to limit what a search result shows, and the three options are frequently confused. They matter even if you want snippets, because the same controls are the only lever you have over the featured snippet: you can decline it, you cannot demand it. Google documents that site owners can opt out of all snippets, including featured snippets, using the robots nosnippet rule or the data-nosnippet HTML attribute, and that when both are present nosnippet takes priority. Google featured snippets documentation (2025).

nosnippet: no snippet at all

The nosnippet rule removes the descriptive text snippet entirely and, with it, any eligibility for a featured snippet. It is the nuclear option, and it is usually wrong for pages that want traffic, because it leaves a bare title link with no description. Use it for pages you want indexed but not summarized, such as paywalled preview pages or pages where a snippet would misrepresent the content.

data-nosnippet: no snippet for a region

The data-nosnippet HTML attribute is surgical. It tells Google not to use the wrapped text for a snippet, so you can protect a specific paragraph, a table, or a block of code from being lifted, while the rest of the page stays eligible. It is the right tool when one part of a page would be a misleading snippet but the page as a whole is fine.

max-snippet: a length limit, and the featured-snippet-only lever

The max-snippet robots rule sets a maximum character length for the snippet, and it has a special property: to block featured snippets only while keeping regular snippets, use the max-snippet robots rule. Google featured snippets documentation (2025). Setting max-snippet to a low value keeps the ordinary snippet short enough that Google will not use it as a featured snippet. This is the only control that removes you from featured snippets while keeping you in the regular results.

Here are the three controls as they appear in the head of a page, so the syntax differences are unambiguous.

<!-- Keep the snippet but cap its length, which also removes featured snippet eligibility -->
<meta name="robots" content="max-snippet:20">

<!-- Remove the snippet entirely, including any featured snippet -->
<meta name="robots" content="nosnippet">

<!-- Allow snippets for the page, but protect one region from being lifted -->
<span data-nosnippet>This sentence must never appear in a snippet.</span>

Measuring AI Visibility: The Generative AI Performance Report vs. Third-Party Scores

AI visibility is the new metric everyone claims to measure, and almost nobody can. Google states that generative AI visibility should be measured with the Search Console Generative AI performance report, and that no third-party tool has access to Google's internal ranking or AI systems. Google AI optimization guide (2025). That sentence is the entire measurement section, and it is worth re-reading.

The Generative AI performance report is ground truth

The report lives in Search Console and shows how your pages perform in AI Overviews and AI Mode: impressions, clicks, and the queries where your content was surfaced. It is the only dataset that comes from Google's own systems, because it is generated from the same internal ranking and AI pipeline that produces the answers. Treat it as the single source of truth for AI visibility.

Third-party scores are estimates, often unanchored

Every third-party "AI visibility score" is an estimate built from sampled queries, a proxy model, or a crawl of AI outputs, because no outside tool can see Google's internal ranking or AI systems. Some of these tools are useful as a directional early warning, but none of them measure what the Search Console report measures, and none should be used to make budget decisions. When a vendor quotes you an AI visibility number, ask what it is actually counting, then check it against the console report.

Run both, but weight the console

The practical setup is to read the Generative AI performance report for truth and keep a third-party tracker only as a supplement for queries and competitors you cannot see in console. Never let a third-party score contradict the console and win the argument. The console is Google's own number; the third-party score is a guess about Google's number.

FAQ and HowTo Rich Results Are Gone for You: The 2023 Deprecations, Dated

If you remember one thing from this guide, make it this: the two rich result types that most ranking guides still tell you to chase were retired for you in 2023. FAQ rich results are now restricted to government and health sites, and HowTo rich results are removed entirely. The timeline below is dated, because the dates are the proof.

Two clarifications prevent the usual panic. First, the changes were not a ranking change: Google stated the FAQ and HowTo changes are not a ranking change and were not added to the Search status dashboard. Google FAQ and HowTo changes (2023). Your pages did not drop because of the deprecation; they simply stopped getting a display treatment. Second, the visible FAQ content on your page still has value for users and for ordinary search relevance. The deprecation killed the rich result, not the content.

The practical cleanup is short. Remove HowTo markup from your templates or leave it, since it does nothing. Keep or add FAQ markup only if you are a government or health site. For everyone else, convert the effort you were spending on FAQ markup into the active types from Step 3, starting with the type that matches each page's primary content.

AI Overviews vs. AI Mode: Why the Links Differ and Why Both Use Query Fan-Out

AI Overviews and AI Mode are often treated as one thing, and they are not. They share a mechanism, query fan-out, and a source, grounding, but they are different products with different models, and the links they show will differ. Google states that AI Mode and AI Overviews may use different models and techniques, so the responses and links they show will vary, and that both may use a query fan-out technique that issues multiple related searches. Google AI features documentation (2025).

Query fan-out is the shared engine

Query fan-out means the system takes your single question and issues several related searches behind the scenes, then assembles an answer from the results of all of them. Google AI optimization guide (2025) describes it alongside grounding as the core of how generative features surface links. The consequence for you is that a page can be cited for a query it does not literally contain, because a fan-out search reached it through a related formulation.

Why the links differ between the two

Different models and techniques pick different supporting pages, so the same question can cite different sources in an AI Overview than in AI Mode. That is expected behavior, not an inconsistency to chase. It also means optimizing for one is optimizing for the other only indirectly: both draw from the same underlying ranking and quality systems, but their surfaces disagree. The correct response is to be a good grounding source broadly, not to chase the specific links of one surface on one day.

Where featured snippets and AI Overviews actually split

Featured snippets and AI Overviews look similar at the top of a page and are frequently confused. The table below is the clean comparison, and the controllability column is the part most people get wrong: neither is opt-in, and only one even has a direct measurement.

The most important row is the measurement row: an AI Overview citation is measurable in Search Console, and a featured snippet is not. That asymmetry is why AI visibility gets a console report while snippet ownership gets a spreadsheet you fill in by hand.

Debugging Checklist: Why Your Rich Result Isn't Showing

When a page has valid markup and no rich result, work this checklist top to bottom. It is ordered by how often each cause is the culprit, from the mechanical to the strategic.

Is the page indexed at all?

Markup on a page Google has not indexed produces nothing. Check URL Inspection, and if the page is not indexed, the problem is crawling or canonicalization, not markup. The technical SEO guide covers the crawl and index layer, which comes before any rich result work.

Did Google fetch the markup?

Look at the rendered HTML Google retrieved, not your browser's copy. A cache, a consent wall, or a tag manager can serve JSON-LD to you and nothing to Googlebot. If the fetched HTML lacks the script tag, fix the serving layer first.

Does the markup still pass the Rich Results Test?

Re-run the test on the live URL. Templates drift, and a markup block that passed three months ago can fail now. Fix every error before looking any further.

Is the markup describing visible content?

Google requires structured data to describe content the user can actually see. Markup that describes hidden, gated, or unrelated content can get the page a manual action rather than a rich result. Align every property with visible text.

Is the type eligible for your site?

Check the eligibility column from Step 3. FAQ markup on a commercial site will validate and still never render. Review markup that is self-serving violates policy. Confirm the type is one your site actually qualifies for.

Is the page simply not the strongest candidate?

Eligibility is not display. Google may show a plain link for a page that passes every test, because a competing page is stronger or because the query does not warrant the enhancement. If everything above is clean and the SERP still shows no rich result, run the query in a clean browser session and use the SERP preview tool to compare your title and snippet against the pages that did win, then move the work from markup to relevance and authority.

Diagnosis Checklist: Why You're Not Cited in AI Overviews Despite Ranking

Ranking on page one and still not being cited in AI Overviews is the most common frustration in 2026. The causes are usually one of these four, and each has a fix.

The page is not snippet-eligible

The entry requirement is snippet eligibility: the page must be indexed and eligible to be shown with a snippet. Google AI features documentation (2025). If you have a nosnippet rule or a snippet-blocking setup, you have removed yourself from the very pool the AI draws from. Check your robots meta for any snippet control.

The answer is not extractable

Grounding lifts a clean, self-contained answer from the page. If your page ranks because of authority but buries the answer in a long narrative, the AI cannot ground a citation in it. Rewrite the opening so the answer sits in the first two sentences in plain prose, the same fix that earns a featured snippet.

The page ranks for a query the AI does not trigger on

AI Overviews only appear when Google's systems determine they are additive to classic Search, and as such they often do not trigger. Google AI features documentation (2025). If the query never triggers an AI Overview, no page is cited, no matter how well it ranks. Check the Generative AI performance report to see whether the query surfaces there at all before you blame the page.

You are measuring the wrong surface

AI Mode and AI Overviews may cite different pages because they use different models and techniques. Google AI features documentation (2025). A page cited in one but not the other is normal. Track the Search Console Generative AI performance report, which aggregates the surfaces, instead of eyeballing one result on one day.

Myth vs. Official Guidance: LLMS.txt, Page Length, Content Chunking, and Other Debunked AEO Tactics

The AEO and GEO vendor market runs on tactics that Google's own documentation explicitly rejects. This section puts each myth next to the official guidance, because the fastest way to stop wasting money is to read Google's answer once.

LLMS.txt and AI text files

Google Search does not use LLMS.txt or other special machine-readable files, AI text files, or Markdown, and creating them neither helps nor harms visibility or rankings in Google Search. Google AI optimization guide (2025). They are not a signal. Build them if some other consumer wants them, but do not expect a ranking effect from Google.

Special AI markup and structured data

Structured data is not required for generative AI search, and there is no special schema.org markup to add, though it remains useful for rich result eligibility. Google AI optimization guide (2025). Structured data still matters, but for rich results, not for AI citations. Keep the two purposes separate.

Page length and content chunking

Google explicitly states there is no ideal page length and no requirement to chunk content for AI, and that rewriting content just for AI systems is unnecessary because AI systems understand synonyms and general meaning. Google AI optimization guide (2025). The "chunk for retrieval" advice is a solution to a problem Google says it does not have.

Manufactured brand mentions

Seeking inauthentic brand mentions across the web is not effective, because Google's generative AI features depend on both core ranking systems and spam-blocking systems. Google AI optimization guide (2025). Paying for mentions that exist only to be seen by an AI is both ineffective and, at scale, a spam risk.

Near-duplicate pages for fan-out queries

Creating many near-duplicate pages to target fan-out query variations primarily to manipulate rankings violates Google's scaled content abuse spam policy and is an ineffective long-term strategy. Google AI optimization guide (2025). This one is not just a waste; it can earn a manual action.

The 2026 SERP Features Reference Table and Launch Checklist

This final section compresses the whole guide into a reference table and a launch checklist you can run against any page. The table is the map; the checklist is the order of operations.

The reference table

The launch checklist

• Benchmark current rich results in Search Console enhancement reports and current features via manual SERP checks.
• Choose one rich result type per page that matches the visible content and passes the eligibility filter.
• Write minimal JSON-LD with only required properties, then add optional ones that exist on the page.
• Validate in the Rich Results Test on the live URL and fix every error.
• Ship, request indexing, and confirm Google fetched the markup in the rendered HTML.
• Track eligibility in the enhancement report and fix invalid pages before warnings.
• For featured snippets, rewrite the opening so the answer is a clean, self-contained sentence that matches the query format.
• For AI Overviews, confirm snippet eligibility and answer extractability, then read the Generative AI performance report.
• Decide snippet opt-outs deliberately: nosnippet for whole-page silence, data-nosnippet for a region, max-snippet to drop the featured snippet only.
• Stop all FAQ and HowTo markup work unless you are a government or health site, and delete any AEO tactic Google's documentation rejects.

That is the whole discipline. Rich results are an opt-in you can verify in an afternoon. Featured snippets and AI Overviews are an outcome you earn by ranking and being extractable. Keep the two columns separate, date your deprecations, and measure each feature with the report that actually exists for it, and you will spend your time on the result types that are still winnable in 2026.