Large swathes of web traffic are already disappearing because AI-powered search now answers questions directly, summarising multiple sources into a single conversational reply. Household-name firms have seen the effect in raw numbers: HubSpot reported a loss of about 140 million visits in one year, and Kipp Bodnar, chief marketing officer at HubSpot, says click-through rates for searches that show AI overviews are about 60 to 70 percent lower. Platforms such as Google have added a dedicated AI-led search experience, and generative tools including ChatGPT, Perplexity and Bing Copilot blend model-generated text with web sources to keep users on the answer screen. That behavioural shift means businesses must add Answer Engine Optimisation, also known as Generative Engine Optimisation, to traditional SEO if they want AI systems to mention their brand.

Brands that don't adapt will lose visits because AI answers replace the click, and that matters because the business impact is both measurable and immediate.

Search platforms are changing how they present information. Google now offers a conversational summary designed to surface an overview and help users dig deeper into a topic. At the same time, generative services such as ChatGPT, Perplexity and Bing Copilot combine model output with web citations. Bottom line: users increasingly receive a ready-made answer at the top of the results page instead of a list of links.

That behavioural change is visible in two places the marketing community watches closely. First, query form has shifted: people now type longer, more conversational prompts when they expect an AI reply, which favours content written in natural language over terse keyword-stuffed pages. Second, the cost of being absent from the summary isn't just fewer visits; it can be fewer qualified leads. Marketers reporting from the field say AEO can increase conversion quality even as raw clicks fall, so the measurement baseline needs to move from visits to outcomes.

2. The three criteria AI uses to decide which sources to cite

Practitioners distil the problem into three interlocking criteria: content alignment, technical machine-readability, and topical credibility. Optimise on all three and you raise the chance an AI answer will include your content. Miss one and you leave the door open for competitors to be cited instead.

Content alignment means publishing answer-first copy that puts the direct response to a likely user question in the opening paragraph. Forbes contributors and industry practitioners urge businesses to build high-authority, clearly structured pages that anticipate long-form conversational queries. Use clear H1 and H2 headers to expose structure, add FAQs and glossaries that map to common prompts, and produce dedicated service or product pages written in natural language that mirror how a person would ask an AI for help.

An everyday worked example helps. For instance, a small rental firm is more likely to be cited in an AI-generated itinerary if it publishes a natural-language article that answers a multi-part query about family holiday plans, activities and local wildlife. The article should open with the concise answer to the likely question, then expand with clearly labelled sections that map to the pieces of the prompt an AI would parse.

Technical machine-readability is the second pillar. Implementing structured data across the site makes content easier for AI-driven systems to parse and attribute.

The recommended types include FAQ schema, product schema with pricing and ratings, organisation schema with canonical business information, and review schema for customer feedback. Use Schema.org as the reference for markup, and ensure product pages expose consistent metadata and internal linking that demonstrates topical clusters.

Topical credibility is the third and often decisive element. AI overviews favour relevance and contextual authority over raw keyword matching, according to SEO specialists. Build high-quality backlinks from recognised industry outlets, accepted directories and respected editorial or trade publications because those links remain signals that downstream models and aggregation systems use to weight reliability. In practice, that means pursuing authoritative coverage, joining relevant specialist directories and encouraging genuine reviews that can be surfaced via review schema.

3. A practical five-step playbook you can deploy

Convert the three criteria into action with a staged approach. The sequence below is the synthesis most UK practitioners follow: small pilots, training, measurement and incremental scaling.

First, audit your existing top-performing pages and identify high-intent topics where a natural-language answer could replace a click. Look for pages that already rank but receive queries that are increasingly conversational. These are the priority topics to pilot.

Mastheads

Second, rewrite or create pages in an Answer-first format. Open with a concise direct answer to the likely user query, then follow with structured detail, embedded FAQs and clearly labelled sections that map to likely AI prompts. Treat headings and Q&A blocks as the machine-readable version of your argument.

Third, add structured data. Use Schema.org to add FAQ schema, product schema, organisation schema and review schema.

Make sure the data exposed in the schema matches the visible content on the page. Explicit Q&A blocks that map to schema make it easier for aggregation systems to attribute material to your site.

Fourth, pursue credibility signals. Seek authoritative backlinks from industry press and trade bodies, list your business in accepted directories and encourage customer reviews that can be surfaced via review schema. The combination of topical relevance and external validation is what models and aggregators use to weight a source’s trustworthiness.

Fifth, pilot, measure and scale. Start with a small set of priority topics, track changes in conversions, lead quality, assisted conversion paths and the specific queries that trigger AI summaries. Iterate on the queries you target: use conversational long-tail prompts in content, measure whether that content is surfaced in AI responses, then expand the content cluster or add supporting schema and backlinks to strengthen the site’s claim to authority.

Execution is organisational as much as technical. UK examples show staged rollouts work. A UK government trial with Microsoft 365 Copilot reported measurable time savings per employee and high retained usage among participants. Organisations such as Lloyds Banking Group scaled Copilot licences only after a staged rollout and training. In professional services, firms including Allen & Overy and several mid-sized legal practices paired AI tools with human specialists, keeping lawyers responsible for review while using bespoke systems for research and drafting.

Conventional metrics will mislead if they remain the only yardstick. Raw clicks and pageviews will often fall even as commercial outcomes improve, so marketers should emphasise outcomes rather than visits alone. Track conversions, lead quality, assisted conversion paths and the queries that trigger AI summaries. Use query-level analytics to discover which conversational prompts are surfacing your pages in AI responses and which prompts are sending the summary to a competitor.

Iterate on content clusters based on those signals. If a targeted answer-first page begins to appear in AI summaries but conversions are low, use the data to adjust the call to action, the supporting FAQs, or the schema. If the page is invisible to AI summaries despite high ranking, focus on credibility work: authoritative backlinks, industry mentions and consistent schema across related pages.

UK practitioners advise training and controlled rollout rather than wholesale replacement. Pilot projects, close training and incremental scaling are the recurrent pattern. That approach reduces risk and lets the organisation learn how to pair AI outputs with human verification, which is especially important in regulated sectors and professional services.

One practical model to follow contains three short cycles. First, pick two priority topics and publish answer-first pages with full schema. Second, run a four to eight week pilot and measure the queries that surface the pages and the downstream conversion quality. Third, expand the programme to the next cluster only after you see improved lead quality or clearer assisted-conversion paths. Repeat the cycle and codify the content templates and schema snippets that worked.

The question that actually decides this is whether your organisation will treat AI-driven answers as a distribution channel that needs nurturing, or as an external threat to be resisted. Treat it as a channel and you gain influence over the way users first see your information.

In short

First, audit and pick priority topics where AI answers would replace a click. Start small.

Second, publish answer-first pages that open with a concise response, use H1 and H2 headings and include FAQs that mirror conversational prompts.

Third, add schema.org markup for FAQs, products, organisation details and reviews so content is machine-readable.

Fourth, build topical credibility with authoritative backlinks, industry press coverage and specialist directories.

Fifth, measure outcomes not visits, iterate on query-level signals and scale the approach that improves lead quality.

Related Articles

Begin with one or two answer-first pages, add FAQ and product schema via schema.org, and track query-level conversions and assisted paths to decide whether to scale.

This article was created with AI assistance.