Chapter 16: AI Search by Industry

Last updated: November 1, 2026.

For an online store, showing up in an AI answer means getting the right price and stock status into a product list. For a machinery manufacturer, it means making the shortlist at the start of a buying process that can take months. For a dental clinic, it means being described correctly on the map and in reviews. A news site has a different worry altogether: if readers get the news from AI, why would they come to the site?

The mechanism behind all of this is the same in every industry. The question is split into searches, pages are retrieved, passages are read and an answer is written. What changes is which data, which sources and which page types carry the most weight. A crawlable site, good content and a trustworthy brand matter in every industry; where you put your budget and time depends on yours.

1. Same mechanism, different weights

AI summaries don’t show up equally often in every industry. Semrush tracked which searches triggered Google’s AI Overviews across more than 10 million keywords in its database throughout 2025 (Semrush). In the November 2025 data, AI Overviews appeared most often for searches in Science (25.96%), Computers and Electronics (17.92%) and People and Society (17.29%). In Real Estate, Shopping and Arts and Entertainment, they appeared for fewer than 3% of keywords.

The same study shows a shift. In October 2024, 91% of searches that triggered an AI Overview were informational. By October 2025 that share had dropped to 57%. Over the same period, the share of commercial searches rose from 8.15% to 18.57%, and transactional searches from 1.98% to 13.94%.

In information-heavy industries the impact starts earlier and runs larger; sites that get their traffic from science, technology and how-to content run into AI summaries the most. Commercial searches are moving into this territory fast, too. The low rate in Shopping doesn’t mean e-commerce is untouched. For product searches Google already has other formats, such as shopping results, and AI works alongside them.

This study isn’t Turkish data. AI Overviews and AI Mode launched in Turkey in February 2026, and the breakdown by industry may look different here. For your own industry, the most reliable data is your own Search Console data and your own prompt measurements. We covered both in Chapters 7 and 8.

2. E-commerce and product data

In e-commerce, the AI answer is often a product list rather than a piece of text. When a user types “a lightweight volleyball shoe for wide feet,” they see products with prices, images, ratings and sellers. That list is fed by your product data, not your blog posts.

Google says so directly in its AI guide: when relevant, AI answers can include product listings, product information and local business information. Products like Merchant Center feeds and Google Business Profile can help your products and services appear in AI answers and other search results (Google Search Central).

The channel is small but growing fast. According to Adobe, traffic from generative AI tools to US retail sites during the 2025 holiday season (November 1 to December 31) rose 693.4% year over year. In the same release, Adobe notes that the user base is still modest (Adobe). The growth rate is big; the volume is still small.

It looks similar across İmza’s clients. Traffic from AI is usually under 10% of the total, but it converts at around 6 to 7%. Given that the general conversion rate in e-commerce is around 1%, that’s no small gap. We went into this in detail in Chapter 1 and Chapter 8.

In e-commerce, the order of priorities:

• Keep product data complete and consistent. Price, stock, color, size, material, shipping time and return terms should match on the page and in the feed. If AI recommends a product as “in stock” and the user sees “sold out” on the page, you’re the one who loses their trust.
• Don’t let the product page be just a spec sheet. A few sentences on who the product is and isn’t for make it easier for AI to match the product to the right need.
• Make category pages work like a buying guide. The constraints users ask about (budget, intended use, size or dimensions) should be reflected on the category page as filters and short descriptions.
• Keep reviews visible and real. AI answers “what do users say?” from the reviews on your site and on marketplaces.

A hypothetical example: a user asks, “Can you recommend a basketball shoe under 5,000 TL with good ankle support that I can also use on artificial turf?” There are four constraints in that question: where it’s used, ankle support, price and product type. If your product page doesn’t mention surface compatibility, ankle support only appears as “high-top” and the price in your feed is out of date, AI can’t know your product fits, even if it does. You don’t need a new blog post for this; completing the existing product page and product data is enough.

We covered product feeds, Merchant Center and buying inside AI in detail in Chapter 15. In e-commerce, AI visibility is a data quality job first and a content job second.

3. In e-commerce, look at revenue

Reading AI’s impact from the traffic chart alone can mislead you. At one of our e-commerce clients, which sells volleyball and basketball shoes, traffic fell 10% while revenue rose 180%. Anyone deciding based on the traffic chart alone would have drawn the wrong conclusion.

AI search moves the user’s research into the conversation. The user who used to land on your blog for “how to choose volleyball shoes” may now get that answer from AI. But the user who comes looking for “that model in size 43” still comes, and is much closer to buying.

So you need to track three things separately. Track traffic to blog, category and product pages separately; growth on product pages can hide a drop in informational content, and the reverse can happen too. Compare the conversion rate of AI-referred visits with organic search. And track revenue and order count: if traffic is falling while revenue rises, there’s no problem; if both are falling, you need a diagnosis.

Marketplaces are a separate issue. Many brands in Turkey sell on marketplaces as well as on their own site, and when AI recommends a product it may link to a marketplace instead of your site. That can look like a loss, but the sale may still be yours. What to watch is that the product information on the marketplace matches what’s on your site.

4. The shortlist in B2B

In B2B, traffic is low, the sales cycle is long and a single sale is worth a lot.

According to Forrester’s 2025 Buyers’ Journey Survey, 94% of business buyers use AI during the buying process, up from 89% a year earlier (Forrester). The same research found that 61% of business buyers use private AI tools provided by their companies. Business buyers are four times as likely as consumers to use Microsoft Copilot and twice as likely to use ChatGPT.

The Copilot part deserves extra attention in B2B. Copilot relies heavily on Bing’s index for its answers. Bing’s share of search in Turkey is very low, which makes it easy to neglect. But if a purchasing manager uses Copilot on a company laptop, whether your site is crawled and indexed cleanly by Bing becomes a sales issue. We covered Bing and Copilot in Chapter 2.

In B2B, what decides things is less the number of visits than whether the right person is visiting. One of our B2B machinery clients gets 30 to 40 visits a day to its site; the machine it sells costs about €20,000. At that scale, a few good inquiries a month can change the year’s sales target. For a company like that, being recommended in an AI answer makes a difference that doesn’t show up on the traffic chart but does show up in the sales report.

In the filling machine example we’ve used throughout the book, the buyer needs to fill 3,000 bottles of a foaming product a day and has a budget of around €25,000. The questions are very specific:

• What nozzle type is used for foaming products?
• How many filling heads does this capacity need?
• Do you provide service in Turkey, and how many days does it take to get spare parts?
• Have you done an installation for a similar manufacturer?

For AI to recommend you in B2B, the answers to these questions have to be written plainly on your site. Moving the answers your sales team gives on the phone onto the site is, I think, the most productive AI work you can do in B2B. We covered finding the questions that matter in Chapter 6, on prompt research.

The decision-maker isn’t one person, either. When a company buys a filling machine, the production manager asks about capacity and product fit, the purchasing manager about price and lead time, the finance manager about return on investment, and the maintenance team about service and spare parts. Each of them asks AI a different question and gets a different answer; you could be recommended in the production manager’s answer and not mentioned at all in the purchasing manager’s. That’s why in B2B the prompt list has to be built from the point of view of every role involved in the buying decision. The persona clusters in Chapter 6 are there for this; skip that step in B2B and half your measurement is blind.

A few more notes. Don’t be afraid to give a price range: you don’t need an exact price, but something like “machines in this capacity range usually cost between X and Y” can get you onto the list for questions with a budget constraint. Application pages such as “Filling for foaming liquids” meet a need the product page can’t. Just don’t open a separate page for every variation; we treat that as a scaled content risk in Chapter 19. Write your references in concrete terms, too. If the customer agrees, a case study with the industry, the capacity and the result is the strongest evidence for AI to say “this company has done this kind of job.” We covered case studies in Chapter 12.

5. Local businesses

For local businesses, the AI answer isn’t mostly drawn from the website. Google Business Profile, maps and reviews weigh at least as much as the site.

A guide by Miriam Ellis of Whitespark, published in May 2026, sums up how AI Mode behaves in local searches (Whitespark). According to the guide, AI Mode pulls ratings, review counts, price information and opening hours from Business Profile and other sources for local results. Ranking high in the classic map results doesn’t guarantee visibility in AI Mode. Two people asking the same question from the same place can see different answers, and reviews have become one of the main sources for AI Mode’s answers to local questions. A GatherUp survey cited in the guide found that 48% of US consumers had used a conversational AI tool to find information about a local business.

A hypothetical example: a dental clinic in İzmir. A user asks, “Can you recommend a dental clinic in Karşıyaka that’s open on weekends, does implants and offers installment payments?” There are four constraints in that question: neighborhood, opening days, type of service and payment option. AI gathers them from the Business Profile, the clinic’s website and reviews. If one piece is missing or the sources contradict each other, the clinic may not make the list. For a local business, half of AI visibility is won off the website.

Priorities for local businesses:

• Keep your Business Profile complete and current: services, opening hours, holiday hours, payment options, photos.
• Describe each service clearly on your site. Instead of one paragraph under “Dental treatments,” write which treatments you offer, who performs them and how the process works.
• Manage your reviews. Ask for reviews and respond to negative ones. Don’t buy reviews; fake reviews break platform rules and, once exposed, wipe out all your credibility.
• Keep your address, phone and opening hours the same on your site, on your Business Profile, on Yandex Maps and in industry directories. Inconsistent information is the most common cause of being described wrongly.

Businesses with more than one location should have a separate page and a separate profile for each one. Creating pages for cities where you have no location is a different matter; it’s one of the best-known examples of the scaled content abuse we’ll cover in Chapter 19.

6. News sites

For news sites, AI search is a different kind of threat than in other industries. In e-commerce, AI sends the user to the product. In news, what the user wants is often the information itself, and AI can hand that over directly.

The Reuters Institute’s Digital News Report, published in June 2026, puts numbers on this (Reuters Institute):

• The share of people using AI chatbots such as ChatGPT, Perplexity and Gemini for news each week rose from 7% to 10% in a year.
• According to Chartbeat data, Google organic search traffic to more than 2,500 sites fell by a third (33%) globally and by 38% in the US between November 2024 and November 2025.
• Publishers expect traffic from search engines to fall by 43% over the next three years.
• The report also notes that AI Mode and AI Overviews are widely assumed to be hurting referral traffic, but that breaking news searches largely fall outside AI Overviews.

That last point is a useful hint for news sites. For breaking news, users still mostly see classic formats like Top Stories. As I read it, this means the traffic a news site loses to AI summaries is concentrated in explainers, background pieces, content that answers “why” and “how,” and the archive. You can check this in your own Search Console data by splitting it by page type.

A few suggestions for news sites:

• Speed is still the biggest advantage. Being indexed first on breaking news is worth more than ever. At one of our news site clients we found that the sitemap was cached for about 10 minutes, so a new story could take up to 10 minutes to appear in the sitemap. We brought that down to seconds. In news, 10 minutes means a competitor gets indexed first.
• Build up loyal readers. In April 2026 Google rolled out Preferred Sources worldwide in all supported languages. Users can tap the star next to Top Stories and pick the publications they like. According to Google, once readers mark a site as a preferred source, they’re twice as likely to click through to it (Google). Reminding readers about this feature is one of the cheapest things you can do today.
• Put original reporting up front. AI can summarize a wire story from anyone, but a story from the field, built on interviews or backed by data, is yours. That’s the news version of the content that can’t be copied from Chapter 12.
• Track Discover separately. Search Console’s AI reports also cover AI features in Discover, and for news sites Discover is often a bigger channel than search.

There’s also a setting you need to decide on. Google added a Search generative AI control to Search Console and made it available to all sites as of August 31, 2026. With this setting you can turn off your site’s appearance in AI Overviews, AI Mode and AI features in Discover. Google says the setting doesn’t affect classic search rankings (Google Search Console Help).

Turning it off may tempt some publishers. It’s better to decide with data: first look at how many impressions and how much traffic you get from AI features; the Search Console report is the place to start. Google notes that even if your site is excluded, similar content from other sites will keep appearing in those features. Opting out doesn’t stop the answer from being given. It only keeps you out of it.

7. Health, finance and law

Health, finance and law are what Google calls YMYL (Your Money or Your Life) topics: subjects that can affect people’s lives, health or money. Google says content on these topics needs to be highly accurate and consistent with well-established expert consensus (Google Search Central).

It’s harder to show up in these industries. AI is more cautious when choosing sources on these topics and leans toward highly trusted ones; government bodies, hospitals, universities and long-established publications are strong here. Getting it wrong also costs more. When a loan’s interest rate, a drug’s dosage or a legal deadline is passed along wrong, the result isn’t comparable to getting a shoe’s color wrong.

Show clearly who wrote the content and who reviewed it. The name, title and area of expertise of the author and of the reviewing expert, the date the content was last updated and its sources should be visible on the page. That’s a trust signal for readers and for AI alike. We covered E-E-A-T in Chapter 2 and expert input in Chapter 12.

Information like interest rates, tax brackets and regulations changes often, and an outdated page can lead AI to give wrong information in your name. In these industries, review your content inventory at least every three months. We’ll build the system for that in the next chapter.

8. Localizing by country and language

For sites that serve more than one country or language, whether the model can find the page that fits the user’s language and country is a separate question.

Google’s guide to multilingual sites sets out the basic rules (Google Search Central):

• Google determines a page’s language from the visible content on the page; it doesn’t use code-level information like the lang attribute or the URL for this.
• It recommends a separate URL for each language version and hreflang markup so the right version is shown.
• It says not to change content based on IP address. If you change content based on language settings, Google may not find all versions.
• It points out that translating only the menus and boilerplate text while leaving the main content in one language makes for a bad user experience.

From an AI point of view, two notes need adding.

Translation isn’t localization. Translating your English page into Turkish doesn’t mean you’ve answered the question a Turkish user is asking. A buyer in Turkey asks about the price in Turkish lira, delivery time from a warehouse in Turkey and service in Turkish cities. A buyer in Germany asks about the CE certificate, the service network in Germany and the price in euros. Same product, two sets of questions in two countries.

Mass-producing pages with machine translation is also a risk. Google’s spam policy lists producing large numbers of pages through automated transformations of other content, such as synonym swapping or translation, among its examples of scaled content abuse (Google Search Central). Using a translation tool isn’t the problem; the problem is publishing translated pages that nobody has checked and that add nothing for the user. We’ll come back to this in detail in Chapter 19.

For the filling machine manufacturer, a buyer in Turkey might be looking for a manufacturer “that installs in Istanbul and accepts installment payments in Turkish lira.” If the same manufacturer exports to the Middle East, the buyer there asks in another language, about another voltage standard, another certificate and the dealer network in the region. Translating the English page into Arabic doesn’t answer those questions; each market needs pages that answer that market’s questions.

Turkish companies that sell abroad should draw up the 20 to 30 questions a buyer in each target country would ask, separately for each country, and measure them in that country’s language. Looking good in one language doesn’t mean you’ll show up in another. We covered measuring platforms separately in Chapter 7; the same logic applies to languages.

9. The picture in Turkey

Google’s AI search is new in Turkey. Google began gradually rolling out AI Mode and AI Overviews in Turkey on February 18, 2026 (Google Turkey Blog); as I write this, it hasn’t even been a year. After AI Overviews launched, none of İmza’s clients saw a serious drop in clicks, only small declines. I don’t expect that to stay the same; trends seen in the US may play out in Turkey a few months or a few years later.

Yandex isn’t small enough to ignore, either. According to StatCounter, in October 2026 Google held 76.85% of the search market in Turkey and Yandex 21.63% (StatCounter). The figure varies with the measurement method, but the ranking has looked similar for years. Yandex launched its AI search for Turkey in December 2024. Called Yazeka, it produces summary answers from multiple sources and cites them (Bloomberg HT). For a site that gets traffic from Turkey, setting up Yandex Webmaster and making sure you aren’t blocking Yandex’s bots is basic hygiene.

It also makes sense to add Yandex to your measurement. In Chapter 7 we talked about measuring platforms one by one; a brand selling to consumers in Turkey can put Yazeka on that list too. Especially for local questions, such as a service in a particular neighborhood or a store in a particular city, Yandex’s map and business data can directly shape the answer. It isn’t known how many people use Yazeka, so size the measurement effort you give it by Yandex’s share of your own traffic.

Local sources are different, too. AI also learns about your brand from what’s written about you on the web, and in Turkey those sources include marketplaces, complaint platforms, sözlük sites (Turkish crowd-sourced forums), forums, local news sites and industry directories. Which of them gets cited on which platform varies by industry and by question; you need to measure this rather than guess. Record which Turkish sites AI cites in your own prompt list. We covered third-party sources in Chapter 14.

The pool of Turkish content is also smaller. That’s a challenge, because when there aren’t enough Turkish sources for a Turkish question, the model may base its answer on sources in other languages, and that answer may not fit conditions in Turkey. It’s also an opportunity: a well-written, current, concrete Turkish source doesn’t face anything like the competition it would in English. I think that’s the biggest advantage brands creating content in Turkey have today.

10. Priority table by industry

The table below sums up what I’ve covered. Use it as a starting point and update it with your own measurements.

Take a second look at the last column. Every industry measures success differently, and in none of them is traffic alone the measure.

11. Start with your own industry

AI search works the same way in every industry, but it doesn’t produce the same result. A generic “GEO project” done without knowing what AI reads in your industry and what users ask usually wastes the budget.

Start with your own row in the table above: in your industry, which source feeds the answer, what’s the first thing to do, and where will you look for success? Once those are clear, the rest of the work falls into order. That order shouldn’t stay a one-off project; it needs to become a cycle that runs again every month, and that’s what Chapter 17 is about.

Checklist

Sources

• Semrush: AI Overviews study (updated December 15, 2025)
• Google Search Central: Optimizing your website for generative AI features on Google Search (last updated July 10, 2026)
• Adobe: Holiday shopping season drove a record $257.8 billion online (January 7, 2026)
• Forrester: B2B Buyers Make Zero-Click Number One (January 22, 2026)
• Whitespark: Guide to Google’s AI Mode for Local Businesses (May 22, 2026)
• Reuters Institute: Digital News Report 2026, Executive Summary (June 16, 2026)
• Google: Preferred Sources language expansion (April 30, 2026)
• Google Search Console Help: Search generative AI control (available to all sites as of August 31, 2026)
• Google Search Central: Creating helpful, reliable, people-first content (last updated October 5, 2026)
• Google Search Central: Managing multi-regional and multilingual sites (last updated December 10, 2025)
• Google Search Central: Spam policies for Google web search (last updated August 28, 2026)
• Google Turkey Blog: Türkiye’de AI Modu ve AI Bakışı dönemi başlıyor (February 18, 2026)
• StatCounter: Search engine market share Turkey (October 2026 data)
• Bloomberg HT: Yandex’ten Türkiye için Yazeka hizmeti (December 17, 2024)

Previous: Chapter 15, Commercial Data and Shopping Agents · Table of contents · Next: Chapter 17, GEO as a Continuous System