Start
Starting audit and score
The list above applied to your own site: a score in four areas and findings in order of priority.
- With Search Console and crawl data
- A first measurement in AI answers
- Preview first, then the final version
SEO · GEO · AEO
Ranking on Google, being cited as a source in ChatGPT, Gemini and Perplexity answers, pages that answer the question directly. We run all three on the same site, with the same measurement sheet.
First I score where things stand today. Then we start with the fix that has the most effect, and measure again every week with the same questions.
- SEOSearch Engine Optimization
- Position in search results. The page gets crawled, gets indexed and comes up near the top for the right search.
- GEOGenerative Engine Optimization
- Being mentioned in AI answers. Engines such as ChatGPT, Gemini and Perplexity describe the brand correctly and cite the page as a source.
- AEOAnswer Engine Optimization
- A direct answer to the question. A quotable page with short answer, definition, table and FAQ blocks.
02.1The work in three parts
Every piece of work ends with a document and a number. Findings, interpretation and estimates are written separately; anything we cannot verify is marked “could not be verified”.
aWhat we do
- Technical auditI find crawling, indexing, speed and schema errors and rank them by impact.
- Search mapWe match category, product and guide pages to the searches they answer, and open a page for searches nothing covers.
- AI answer trackingWe fix the questions the brand wants to be mentioned in and ask them to ChatGPT, Gemini, Perplexity and Copilot every week.
- Quotable textWe write short answer, definition, table and FAQ blocks and mark them up with JSON-LD. We add llms.txt too, as a harmless extra whose effect is unproven.
bWhat we deliver
- Starting score and list of findingsEach finding with its evidence, impact and fix.
- A roadmap in order of priorityWhich fix comes first, which comes later, and why.
- Fixes in placeOr a task list a developer can pick up and do directly.
- Question set and tracking sheetWhich question the brand was mentioned in and which page was cited, week by week.
cHow we measure
- Search ConsoleImpressions, clicks, average position, number of indexed pages.
- AI answersMention rate on the fixed question set, rate of correct descriptions, number of pages cited as sources.
- TechnicalCore Web Vitals field data, crawl errors, schema validation.
- DatesEvery measurement is dated; the value from before the work started always sits next to it.
02.2Visibility checklist
Some of the items checked in the starting audit. Click an item to see what is checked, how it is checked and the first step.
Starting audit
An e-commerce site, first week
0/100
score · overall 0
Done 0Partial 0Missing 0
-
Important pages missing from the index
Missing · high impact
- What we check
- Whether the category and product pages that bring in sales are in Google’s index; pages that land in the “Crawled, currently not indexed” list.
- How we check
- The Search Console pages report is compared with the URLs in the sitemap.
- First step
- Duplicate or thin pages are merged, given internal links and submitted for recrawling.
-
One correct canonical URL
Partial · high impact
- What we check
- Which main page the filter, sort and campaign parameter URLs point to.
- How we check
- The site is crawled with a crawler tool; pages whose canonical tag does not point to themselves are listed.
- First step
- Parameter URLs are pointed to the main page and taken out of the sitemap.
-
Mobile speed: LCP, INP, CLS
Partial · medium impact
- What we check
- Largest content load time, responsiveness and layout shift in real-user data.
- How we check
- Search Console Core Web Vitals and PageSpeed Insights field data; a lab score alone is not enough.
- First step
- The template with the most traffic (often the product page) is fixed first.
-
robots.txt and sitemap agree
Done · medium impact
- What we check
- Every URL in the sitemap returns 200 and is not blocked by robots.txt.
- How we check
- The sitemap is downloaded and its URLs are requested in bulk.
- First step
- No problem in this example; we check again once a month.
-
Redirect chains and 404s
Done · low impact
- What we check
- Redirects with two or more hops, broken internal links.
- How we check
- 3xx chains and 4xx targets are filtered from the crawl output.
- First step
- Chains are cut to a single hop; internal links are updated to the final URL.
-
Duplicate titles and meta descriptions
Partial · medium impact
- What we check
- Groups of pages that share the same title or description.
- How we check
- The title and description columns in the crawl output are grouped.
- First step
- The template gets the page’s own details (category name, product attribute).
-
Each category answers one search intent
Partial · high impact
- What we check
- Whether the search a category page targets matches the page’s heading, text and product list.
- How we check
- Search Console queries are pulled page by page; two pages competing for the same query are flagged.
- First step
- One of the competing pages is chosen and the other links to it.
-
Internal links: guide to category, category to product
Missing · high impact
- What we check
- Whether guides link to the right category and categories link to their featured products.
- How we check
- Number of internal links each page receives, from the crawl output.
- First step
- Revenue pages with few links go to the top of the list.
-
Thin and duplicate content
Partial · medium impact
- What we check
- Products with the manufacturer’s description copied word for word, category texts a few sentences long.
- How we check
- Text similarity scan and word count distribution.
- First step
- Original descriptions are written first for the pages with the most impressions.
-
One H1, headings in order
Done · low impact
- What we check
- One H1 on every page and an H2, H3 order that does not skip levels.
- How we check
- Heading columns in the crawl output.
- First step
- No problem in this example.
-
Image alt text
Missing · low impact
- What we check
- Descriptive alt text on product and guide images.
- How we check
- Empty alt text and alt text that is just a file name are listed.
- First step
- An alt text template is built from the product name and its distinguishing attribute.
-
Author and update date are visible
Missing · medium impact
- What we check
- Who wrote each guide and when it was last updated.
- How we check
- The page and its schema are read together.
- First step
- A visible line is added and the same details go into the Article schema.
-
A fixed question set is in place
Done · high impact
- What we check
- The questions the brand wants to be mentioned in: category questions, comparison questions, “where to buy” questions.
- How we check
- The same question list is put to ChatGPT, Gemini, Perplexity and Copilot every week, the same way.
- First step
- The list stays unchanged; otherwise the weeks cannot be compared.
-
Answers describe the brand correctly
Partial · high impact
- What we check
- Which line of business and which features the AI answer links the brand to; outdated or wrong information.
- How we check
- Answers are saved; sentences that describe the brand wrongly are flagged together with their source page.
- First step
- The page the wrong information comes from is fixed; one clear definition sentence is written on the brand’s own site.
-
We know which pages get cited
Missing · high impact
- What we check
- The URLs cited as sources in answers: the brand’s own site, competitors, forums, comparison sites.
- How we check
- Source links are collected for every question and counted by domain.
- First step
- The kind of competitor page that gets cited (table, guide, FAQ) is matched on the brand’s own site.
-
llms.txt and a plain-text version
Missing · low impact
- What we check
- An /llms.txt file that sums up what the site is and lists its important pages.
- How we check
- Whether the file exists and whether its links work.
- First step
- A short index is written from the category and guide pages. A harmless extra: Google does not use it, and its effect on other engines is unproven.
-
An access decision for AI crawlers
Partial · medium impact
- What we check
- How crawlers such as GPTBot, PerplexityBot, Google-Extended and ClaudeBot are handled in robots.txt and in the firewall.
- How we check
- robots.txt, CDN rules and server logs are read together.
- First step
- Which crawlers are allowed is decided on purpose and written down.
-
Mentions outside the site
Missing · medium impact
- What we check
- Whether the brand is mentioned on guide, forum, comparison and news pages.
- How we check
- The brand name is searched for on the outside pages cited for the question set.
- First step
- For cited pages that do not mention the brand, a way in is worked out one by one.
-
A short answer at the top of the page
Partial · high impact
- What we check
- Whether the answer to the question is given in the page’s first paragraph, in 40-60 words.
- How we check
- The first paragraph of each guide is read side by side with its question.
- First step
- Filler in the intro is cut and the answer moves to the first sentence.
-
FAQ block and FAQPage schema
Missing · medium impact
- What we check
- An FAQ built from real customer questions, with the same text in the schema.
- How we check
- The Rich Results Test and the page text are compared.
- First step
- The questions the support and sales teams hear most often are collected.
-
Comparison tables in plain HTML
Done · medium impact
- What we check
- Whether tables arrive as readable HTML tables as soon as the page opens.
- How we check
- The page is opened with scripts turned off.
- First step
- No problem in this example.
-
Definition sentence
Partial · medium impact
- What we check
- Whether key terms are defined in one sentence of the form “X is a … that …”.
- How we check
- In each guide we look at the sentence where the term first appears.
- First step
- Every guide gets a definition sentence, and a short table where needed.
-
Organization, Product, BreadcrumbList schema
Partial · medium impact
- What we check
- Whether the schema matches the visible information on the page and has no errors.
- How we check
- The Schema validator and the Search Console enhancements report.
- First step
- Faulty fields are fixed; information that is not visible is not written into the schema.
-
Figures are dated and sourced
Missing · low impact
- What we check
- Whether prices, rates and statistics on the page come with a date and a source.
- How we check
- Figures in the guides are flagged one by one.
- First step
- A figure whose source cannot be found is removed.
Score: done 1, partial 0.5, missing 0; the average of all items.A made-up example; it does not belong to any client.
02.3Related packages
The three packages that most often go with this work. We set the scope together, based on the kind and size of the site.
Research
Competitor research
Which kinds of pages the competitors that stand out in the same searches and the same AI questions use to stand out.
- Competitor pages that get cited
- Searches a competitor covers and you leave empty
Continuity
Regular reporting
Regular measurement with the same question set and the same Search Console view; the value from before the work always next to it.
- Weekly AI answer table
- Monthly search summary
- A dated log of the fixes made
Send the site and the searches you track.
Write the address and a few searches or questions you care about, and I will take a first look from there.