GEO makes the entity clear enough to cite.

Generative engine optimization is the work of making a company, a product and a page identifiable, then writing so a system can quote the page without inventing the source. We measure AI referrals as their own report, not as a slice of ordinary search.

Check your site free

What GEO is, on a site we control

GEO is not a second copy of SEO with a new name. Ordinary search still has to crawl and index the page. GEO adds a different question: when a system writes an answer, can it tell who you are, and does the page contain a sentence it can cite?

The entity is the stable identity. Same organization name, same URL, same profiles, the same description of what you do. If LinkedIn says one thing, the homepage another, and the footer a third, a model has nothing solid to attach a citation to. We make those match before we write more pages.

SEO, GEO and AEO, togetherVideojam search resultsTechnical SEOAnswer engine optimizationAI search readiness checker

How Audacian describes itself to machinesThe entity graph behind this site. One Organization node in JSON-LD, linked to the people who run it, the profiles listed in sameAs, the llms.txt files and the schema on each page. Every link is a real field you can read in the page source.
  • Organization → Person: founder and employee
  • Organization → sameAs: LinkedIn, Crunchbase, Clutch, YouTube
  • Organization → llms.txt: every URL with its summary
  • Organization → Page schema: Service, FAQPage, VideoObject
  • Organization → WebSite: publisher → Organization

How we do it

We write the entity down once and reuse it. On this site that is an Organization record with a stable @id, the same name on every page, and sameAs links only where we actually have a profile. Founders are Person records linked back to that organization. Service pages point at the organization. The markup matches the sentence a person can read.

We publish llms.txt and a longer llms-full.txt so a crawler can see the site map in plain text. robots.txt names the crawlers we allow, including the ones used by answer engines, and points at the sitemap. We do not block a crawler and then wonder why we are not cited.

Citation-shaped writing is a page that answers in the first sentences, names the thing it is about, and gives the figure with its date and its source. Measuring it means a separate look at AI features and at referral logs. We do not treat an AI impression as a click, and we do not subtract it from web search.

  • One Organization, the same @id, sameAs only for real profiles.
  • llms.txt and llms-full.txt generated from the pages we ship.
  • robots.txt that names the AI crawlers and links the sitemap.
  • A page lead that states the answer, then the evidence.
  • AI impressions reported beside web search, not inside it.

Videojam’s AI-features export

For 1 August to 31 October 2026, Videojam’s web search report shows 12,486 impressions and 272 clicks. The AI-features export for the same dates is a different file. It shows 749 impressions. It has no click column and no queries. 729 of those impressions are dated from 23 September onward.

The daily rows in that AI export add up to 778, which is not 749. We quote 749 because that is the export’s total, and we say so. We do not “reconcile” the two by hiding the gap. We also do not treat 749 as a subset of the 12,486. They are different reports. Subtracting one from the other would invent a number the exports do not contain.

What we can say is narrower. Web search and the AI report both move after 23 September, the day after the crawl-path work. That is the same correlation as on the technical page. It is not a measurement of citations, and it is not a count of how often a model named Videojam.

Web search and AI features, 1 Aug–31 Oct 2026Web search: 12,486 impressions and 272 clicks. The AI-features export is a separate report: 749 impressions, no click column, and 729 of those impressions from 23 Sep onward. Do not subtract 749 from 12,486. The daily rows in that export sum to 778, so the figure we quote is the export total.
SeriesCount
Web search impressions12,486
AI-features impressions749
Clicks272 (Web search impressions)
Videojam search performance1 Aug–13 Nov 2026 · Google Search Console, web search
Impressions
17,469
Clicks
342
CTR
1.96%
Avg. position(lower is better)
21.2
vs previous 28 days
The export starts on 1 August, so there is no earlier period to compare.

Impressions (left axis)Clicks (right axis)22 Sept: Crawl-path and event-hub work shipped

002501050020750301,000401 Aug1 Sept1 Oct1 Nov22 Sept

The rise starts the next day. A correlation, not proof of cause.

Show the numbers as a table
DateImpressionsClicksAvg. position
1 Aug 202619041.9
2 Aug 202624035.3
3 Aug 202614039.3
4 Aug 202610038.8
5 Aug 20267036.3
6 Aug 202614033.6
7 Aug 20267031.4
8 Aug 20266037.7
9 Aug 202629327.1
10 Aug 202619131.6
11 Aug 202621036.0
12 Aug 202612043.7
13 Aug 202620142.9
14 Aug 202614433.1
15 Aug 202619056.1
16 Aug 202622059.4
17 Aug 202613057.1
18 Aug 202620047.9
19 Aug 202611060.8
20 Aug 202615142.9
21 Aug 202612337.3
22 Aug 202618068.4
23 Aug 202622052.5
24 Aug 202619339.2
25 Aug 202627049.1
26 Aug 202626050.0
27 Aug 202629057.4
28 Aug 202634058.9
29 Aug 202631052.7
30 Aug 202634154.4
31 Aug 202629054.8
1 Sept 202613036.8
2 Sept 202618225.9
3 Sept 202613140.5
4 Sept 202616141.2
5 Sept 202614140.0
6 Sept 202624041.3
7 Sept 202620046.5
8 Sept 202616042.8
9 Sept 202627053.4
10 Sept 202623024.4
11 Sept 202626047.7
12 Sept 202626152.4
13 Sept 202622146.5
14 Sept 202626155.6
15 Sept 202621054.0
16 Sept 202624047.8
17 Sept 202612129.2
18 Sept 202613046.8
19 Sept 202613050.0
20 Sept 202611148.8
21 Sept 202610056.3
22 Sept 202613059.2
23 Sept 2026133422.8
24 Sept 2026185221.8
25 Sept 2026206616.7
26 Sept 2026194416.9
27 Sept 2026163414.4
28 Sept 2026167412.2
29 Sept 2026222718.8
30 Sept 2026193719.5
1 Oct 2026196522.7
2 Oct 2026218816.1
3 Oct 2026214418.9
4 Oct 2026255119.4
5 Oct 2026223315.6
6 Oct 2026237516.5
7 Oct 2026244319.9
8 Oct 2026232216.7
9 Oct 2026253714.1
10 Oct 2026231114.5
11 Oct 2026248320.3
12 Oct 2026269415.6
13 Oct 2026306518.6
14 Oct 2026359717.9
15 Oct 2026387818.7
16 Oct 2026441419.2
17 Oct 2026389522.6
18 Oct 2026357619.9
19 Oct 2026408719.1
20 Oct 2026332820.7
21 Oct 2026404719.1
22 Oct 20264561520.2
23 Oct 20265082420.2
24 Oct 20264581517.0
25 Oct 2026356724.3
26 Oct 20263481017.6
27 Oct 2026375621.4
28 Oct 2026335921.2
29 Oct 2026344318.4
30 Oct 20263161021.2
31 Oct 2026326520.4
1 Nov 2026344718.3
2 Nov 2026317421.0
3 Nov 2026359722.5
4 Nov 2026343622.2
5 Nov 2026335621.3
6 Nov 2026332421.1
7 Nov 2026356224.6
8 Nov 2026328324.3
9 Nov 2026307823.9
10 Nov 2026327723.6
11 Nov 2026347620.0
12 Nov 2026397020.6
13 Nov 20268911018.1

Get a free 20-min SEO and AI-search review of your site

We open your site and Search Console with you and point at what is holding it back.

What we do on audacian.eu

This site is prerendered, so the HTML a crawler fetches already contains the text. Each indexable URL has a canonical, a title, a description and hreflang for English, Spanish, Romanian and French, plus x-default. The sitemap lastmod comes from the last git change of that page’s source.

Organization sameAs points at our profiles on LinkedIn, Crunchbase, Clutch and YouTube. We add a profile to sameAs only once it is live and matches our details. The founder is a Person linked to the organization. There is no personal LinkedIn on that Person, because we have not put one in the source.

Production deploys ping IndexNow. robots.txt allows Googlebot, Bingbot, GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and Google-Extended. That is a policy choice you can read in the file. It is not a claim that any of those systems cite us.

What GEO does not do

GEO does not make a model mention you. A model cites what it was trained on and what it can fetch. We can make the page fetchable, the entity consistent and the sentence quotable. We cannot make the mention happen, and we will not sell a forecast of how often it will.

We also will not present an industry percentage for “AI Overviews” or “zero-click searches”. Those numbers change, and the ones that circulate are often unsourced. The only AI figure on this page is Videojam’s 749 impressions, with the caveats above.

On a client site the same rule holds. We will publish the organization name, the URL and the profiles you actually have, allow the crawlers you want allowed, and put a quotable sentence at the top of the pages that should be cited. We will not add a profile you do not have in order to fill sameAs, and we will not report an AI impression as a visit. If your analytics show no AI referral yet, that is the measurement. It is not a reason to invent one. The two bars on this page are the two Videojam exports for 1 August to 31 October 2026. The table under them repeats the same two counts, plus the 272 web clicks, which the AI export does not have.

We will not promise a ranking, a traffic number, or a citation in an AI answer. You get the audit, the changes, and a second read of the same Search Console export, so the before and after sit on one page. If the indexed count does not move, we say that too.

GEO questions

What is GEO?

Generative engine optimization is making an organization and its pages identifiable and quotable. It sits on top of ordinary indexation. It covers a consistent entity, crawler access, llms.txt, and pages that state the answer in the first sentences.

How is an AI impression different from a search impression?

On Videojam they come from different Search Console exports. Web search for 1 August–31 October 2026 is 12,486 impressions and 272 clicks. The AI-features export is 749 impressions, with no clicks and no queries. We do not subtract one from the other.

Why do the AI rows not match 749?

The daily rows in that export sum to 778. The export total is 749. We quote the total and we mention the row sum, so the gap is visible. We do not pick the larger number to make the chart look stronger.

What goes in llms.txt?

A short plain-text map of the site: who we are, the services, and the translated URLs. llms-full.txt adds each indexable page’s title and description. Both are generated at build from the same routes as the sitemap.

Which profiles do you mark up?

Only profiles that are live and match our details. Audacian’s Organization sameAs is LinkedIn, Crunchbase, Clutch and YouTube. The founder Person has no personal LinkedIn until we have one to publish.

Can you guarantee a citation in ChatGPT or Perplexity?

No. We can show that the crawlers are allowed, that the entity is consistent, and that the page answers in its first sentences. A mention is the model’s choice.

Tell us what you're launching.

Book a 20-minute intro call, or email a few lines about what you're launching. We'll reply with a rough plan and a price range, whether or not we end up working together.