An answer engine does not read a page from top to bottom. It looks for a passage that answers the question and checks it against what it knows about the topic from other sources. A page written around one keyword rarely passes that test, and it costs clicks: Ahrefs compared 300,000 keywords and found that when an AI Overview sits above the results, the page in first position gets a 58% lower click-through rate on average (desktop, December 2025). These are the three causes we see most often in content audits.
The definition sits in the middle of the third paragraph and the steps are described in sentences. There is nothing for the engine to lift, so it quotes a page that gave the same answer in two sentences.
Repeating a keyword tells an engine nothing new about the topic. In the GEO study that tactic scored worse than leaving the text unchanged.
An article about whole-bean coffee that never mentions roast level or brewing method looks shallow to a model. A competitor that covers them gets the citation.
In our reading of 143 pages cited by four AI engines for six questions about choosing an agency in Poland, articles made up 54% and service pages and agency directory listings 24%. So we work on both, and each type has its own form and its own price.
| Page type | What we change | When it makes sense |
|---|---|---|
| How-to and guide articles | Definition up front, steps as a list, questions with answers, sources and dates | Customers ask how something works or how to choose |
| Old posts with traffic | Updated data, new concepts, a split into passages, a new review date | The post ranks in Google, but engines skip it |
| Category descriptions | A description matched to buyers' queries by comparing vectors, plus the questions buyers ask | The store has categories with traffic and a short or copied description |
| Ranking articles | Criteria up front, a table, a short reason for each entry | Customers ask the engines which product or company to choose |
| Service pages | A clear service name, scope, price and answers to pre-purchase questions | An engine describes your offer vaguely or not at all |
We start with what you already have. Fixing a good post costs less than writing a new one, and it often works sooner.
first week
We collect the questions your customers ask in Google and in AI engines and match them to pages of the site. A question with no page is a topic for a new article.
first two weeks
For each question we look for the passage on your site that answers it. We check which posts and categories have traffic, what they lack and which pages compete for the same question.
first two weeks
For each page we assign the scope of work and a cost: a fix, a rewrite or a new article. You approve the list before we start writing.
as planned
We write, add the structured data and hand over each text with a note of what changed. We publish it ourselves or pass on the finished files.
after 4-8 weeks
We check whether the reworked pages appear in answers to the questions on the list and how traffic from Google changed. What we learn goes into the next batch.
We ask the same questions before and after the change. Five indicators show whether the text reaches the answers and whether the answers bring people to the site.
| Indicator | Where the data comes from | What it tells you |
|---|---|---|
| Citations of the reworked pages | The source lists under answers in ChatGPT, Gemini, Perplexity and AI Overviews | Whether an engine names your page as a source |
| Positions and impressions in Google | Google Search Console, queries assigned to the page | Whether the page gained in classic search |
| Traffic from AI engines | GA4, referrals from chatgpt.com, perplexity.ai and gemini.google.com | Whether the answers bring people to the site |
| Topic coverage | Comparing the concepts in the text with those on pages the engines cite | What the article still lacks |
| Pages competing for the same question | Search Console, several addresses for one query | Whether the site is taking traffic from itself |
Every month we ask four AI engines the questions that customers in Poland ask when choosing a marketing agency. In the measurement of 21 September 2026 the engines most often reached for articles, but almost one citation in four went to a service page or an agency directory listing.
of cited pages were articles: rankings, guides, round-ups and answers to questions
were service pages and agency directory listings, so a service description can reach the answer too
is the most extra visibility that quotations, statistics and sources in a text gave in the GEO study
Ratings of JustIdea from three public profiles, including Clutch, which verifies each reviewer in a conversation before publishing. It is also an example of content that engines like to quote: specific, signed and checkable.
weighted average of 224 reviews across three platforms. Clients gave them, and each site links to the profile with the full text of every review.
Read the reviewsWe list the partner statuses we actually hold and say what each one is useful for in answer engine optimization.
Answers in Google are lifted from its live index, so a fix on a page shows up there sooner than in the memory of a language model. As a Premier Partner we also run the ads on the same results page, so for your queries we see both the paid results and the answer above them.
Copilot answers from Microsoft's own index, a second place where a reworked page has to be found. We run Microsoft Advertising campaigns as an Elite Partner and check how Bing indexes your page in Bing Webmaster Tools, separately from Google.
Part of what engines read about a company sits in its social profiles. We run Meta ads as a Business Partner and keep the company data in your profiles identical to the website, which supports the consistent company entity that our glossary of concepts builds.
The price does not differ from ordinary copy, because we do not add a surcharge for the word AI in the name of a service. Text is billed per 1,000 characters, categories and rankings per piece, with no retainer.
Content and optimization for AI at the price of classic copy.
Optimizing product category descriptions with semantic vectors (cosine similarity), to match them better to the requirements of search engines and LLMs.
Comparison articles (top 5 or top 10 products) on your blog or on external portals.
The call is led by a new business specialist. When we get into the details of an account, the specialist for that channel joins in. We reply within one business day.
Before the meeting we review your website, your visibility and what your campaigns show from the outside. We will not open with “so, what does your company do?”.
The reply comes from the same new business specialist who will run the call. No qualification form and no call from an unknown number.
We go through your numbers and your questions. If it turns out we are not the right fit, we will tell you straight away.
