Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
Why This Category Went First
A recipe is a list of ingredients and a sequence of steps. That compresses into a generated answer with almost nothing lost, which is not true of a product comparison, a diagnosis, or an argument. Food creators were not unlucky; they publish the most summarisable content on the internet.
The consequences were severe and widely reported through 2025, with many creators describing losses across a wide range. Treat the specific percentages circulating with caution, since they come from self-reported surveys and vendor analyses rather than disclosed datasets, but the direction is not in dispute and the pain is real.
The structural problem underneath is that food blogging monetises through page views. When the answer is delivered without the visit, the revenue model breaks even where the content was genuinely the source. That is why this category's response has been louder than others, and why the platform response arrived here first.
What Changed in March 2026
Google updated AI Mode specifically for recipes, aimed at the problem of answers assembled from several sources without connecting people to the creators behind them. Robby Stein of Google said they had heard the feedback on recipe results and were making updates to better connect people with recipe creators on the web.
The practically important detail is where the surfaced information comes from. The experience draws on Recipe structured data, including preparation and cook times, the author, aggregate ratings, and a hero image. Those are fields many creators fill partially or leave to a plugin default.
That produces an unusual situation. Across most of the web the set of structured data that earns anything visible has been shrinking, as covered in structured data in 2026. Recipe went the other way: it kept its rich result and acquired a second job feeding an AI surface. Completing that markup is now one of the few schema tasks with a clear current payoff.
Getting the Markup Genuinely Complete
Most recipe plugins emit valid markup with substantial gaps. Valid is not complete, and the gaps are exactly the fields being surfaced. Open your own markup and check rather than trusting a green validation.
| Field | Commonly | Why it matters now |
|---|---|---|
| prepTime, cookTime, totalTime | Partially filled or omitted | Surfaced directly; a filter people apply |
| author | Site name rather than a person | The attribution link; see author entities |
| aggregateRating | Absent when no rating widget exists | Surfaced, and only valid from genuine user ratings |
| image | One image, small or wrongly cropped | Multiple aspect ratios give a usable hero image |
| recipeYield, nutrition | Frequently skipped | Answers questions people search directly |
| recipeIngredient, recipeInstructions | Present but unstructured | Properly separated steps are extractable individually |
One rule constrains all of this: markup must describe what the page visibly shows. Ratings must come from real users, times must match the printed recipe, and nutrition must be what you actually calculated. Fabricated ratings are the most tempting shortcut in this category and the one with the clearest policy exposure.
The Preamble Problem, Honestly
The long story above the recipe is the most complained-about pattern on the web, and it exists to increase ad impressions rather than to serve readers. There has never been good evidence that it was what made these pages rank.
The usual defence, that context and personal narrative satisfy quality expectations, confuses two things. Original photography, tested variations, substitution advice, and hard-won notes about what goes wrong are genuinely valuable and are exactly what a generated summary cannot replicate. Six paragraphs about a holiday before the ingredient list are not that.
Put the recipe card near the top, keep a short piece of context, and move the substantial material, the variations, the failure modes, the equipment notes, below it where the people who want it will read it. That structure also gives retrieval discrete, self-contained sections rather than one long block, which is the property described in query fan-out.
The Part That Is Not an SEO Fix
A business that monetises page views is exposed to any change that reduces page views, and no amount of markup changes that. The creators handling this best moved the relationship off search without abandoning search as discovery.
Concretely that means an owned channel people opt into, where the recipe arrives without an intermediary deciding whether to show it. It also means income that does not scale purely with impressions: paid subscriptions, a cookbook, classes, or products. Search then becomes the top of a funnel rather than the whole business.
The uncomfortable framing is that this was always the exposure and the traffic simply made it invisible. The general case is set out in what you still own when search moves, including the finding that AI referrals did not replace what search stopped sending, and the creator-specific version is in SEO for content creators.
Questions People Ask About Recipe SEO
- Why were food blogs hit so hard by AI answers?
Because a recipe is the most summarisable content there is. Ingredients and steps compress into a generated answer with almost nothing lost, unlike a comparison or an opinion. Food and drink was among the fastest-growing categories for AI answer triggers, and many creators reported severe traffic declines through 2025.
- What did Google change for recipes in March 2026?
Google updated AI Mode specifically to address recipes being assembled from multiple sources without properly connecting people to the creators behind them. Google's Robby Stein said they had heard the feedback and were making updates to better connect people with recipe creators on the web, with the experience surfacing details drawn from Recipe structured data.
- Does Recipe schema still matter?
More than it used to, which is unusual given how many structured data types have been retired. Recipe remains a supported rich result type, and the details AI Mode surfaces for recipes are drawn from that markup: times, author, ratings, and images. Incomplete markup means less to surface, which is a direct cost.
- Should I remove the story above my recipe?
Shorten it and stop pretending it is there for readers. The long preamble exists to increase ad impressions and it is the single most complained-about pattern in the category. A recipe card near the top with the context below serves people better, and there is no evidence the preamble was ever what made these pages rank.
- What should food creators do about declining search traffic?
Treat search as discovery and build something that does not depend on it. The creators weathering this best have a direct channel, an email list or an app, and use search to fill it rather than to monetise it. That is a business model change rather than an SEO fix, and no amount of optimisation substitutes for it.
