AIShortlist documentation

What do you need before installing?

Your keys, your account

Measurements run through your own provider accounts. Keys are stored in your own database and are sent to nothing but the provider they belong to. This add-on has no server of its own and phones home to nobody.

The cost is therefore yours: the one-time licence does not include API usage. One run asks each active question to each selected engine once: ten questions across three engines is thirty calls. Within the reuse window a repeated run costs nothing at all.

As a reference, in our own measurement on 24 September 2026 one question cost about $0.013 on Perplexity (sonar-pro) and about $0.03 to $0.04 on OpenAI (gpt-5-search-api). We did not measure Gemini separately. The amount depends on the model, the length of the answer and your provider's current pricing.

How do you install the add-on?

1. Copy the aishortlist folder into system/user/addons/.
2. In the control panel go to Developer Tools → Add-Ons and press Install next to AIShortlist.
3. Open Settings, enter your brand name and at least one API key.
4. Open Questions and write the questions your buyers would ask.
5. Press Run measurement now.

The first run takes a while: search-grade models are slow, and the add-on deliberately pauses between calls so the providers’ per-minute token limits are not tripped. Three questions across three engines takes roughly three minutes.

Settings reference

Your brand

Brand name is the name you want to be recommended by. Other spellings takes one variant per line. Engines write names inconsistently, and adding the variants you have seen keeps the score honest. Leave it empty to match on the brand name alone.

Products to track takes one per line as Name | keyword, keyword. Without the pipe, the name is used as the keyword. A product counts as mentioned when any of its keywords appears in the answer.

API keys

A saved key is never displayed back to you. Leave a field empty to keep the key you already have; type a new one to replace it. Leave a key empty entirely and that engine is skipped.

Engines and limits

Scheduled measurement

ExpressionEngine 6 and 7

These ship a CLI, so a normal cron entry is enough:

0 6 * * 1  cd /path/to/site && php system/ee/eecli.php aishortlist:olc

ExpressionEngine 5, or any host without shell access

Use the URL shown at the bottom of the Settings screen. It carries a token that is generated for your site:

0 6 * * 1  curl -s "https://example.com/?ACT=12&jeton=..."

Weekly is usually the right frequency. Answer engines do not change their recommendations hour to hour, and a weekly line is easier to read than a daily one.

Template tags

Every figure in the control panel is also available in templates, so the numbers can go onto a client report or a dashboard page.

Score

{exp:aishortlist:score}
    We are named in {mentioned} of {measured} AI answers ({visibility}%).
    {if best_position}Best position: {best_position}{/if}
    {if no_results}No measurement has run yet.{/if}
    Last measured {run_date format="%Y-%m-%d"}.
{/exp:aishortlist:score}

Competitors

{exp:aishortlist:competitors limit="10"}
    {count}. {name}: named in {mentions} answers
{/exp:aishortlist:competitors}

Results

{exp:aishortlist:results engine="openai"}
    {question} → {engine_name}
    {if mentioned}named us at position {position}{if:else}not named{/if}
    {evidence}
{/exp:aishortlist:results}

engine is optional and accepts openai, gemini or perplexity. Omit it for every result. All three tags read the most recent run.

Which models are used, and what do the results represent?

In version 1.0.0, when the model fields are empty the add-on uses these defaults. All three answer with web search.

The results represent the API answers of the provider and model you selected. Each question is asked once per engine per run, with a short instruction added that asks for current web sources and a numbered list. The answer a logged-in user sees in the provider's own app can differ. If you change a model, the series before and after the change is no longer directly comparable.

A failed call is tried three times, then recorded as an error. It is left out of the score: the score is the share of measured answers that name your brand.

How does the add-on differ from the AIShortlist dashboard?

The dashboard is described at aishortlist.app/en/features/; how it measures is on the methodology page.

What does the add-on store?

Four tables, all prefixed exp_aishortlist_: settings, questions, the raw answers, and the scores derived from them. Uninstalling the add-on drops all four.

Raw answers are kept on purpose. Scoring an answer costs nothing, so when you change your brand spellings the whole history can be re-scored without asking a single question again.

Multi-site installs

Every row carries a site ID. On an MSM install each site keeps its own brand, its own questions and its own measurements, and they never mix.

What if something does not work?

A run reports errors for one engine

The message carries the provider’s own response. insufficient_quota or a billing message means that account is out of credit; the add-on stops retrying that engine for the rest of the run rather than wasting time. Repeated 429s mean the delay between calls is too short, so raise it.

The measurement says 0% and you expected better

Check the evidence column before assuming a bug. If the engines genuinely do not name you, that is the finding the tool exists to deliver. If they do name you but spell it differently, add that spelling under Other spellings and run again. Stored answers are re-scored, so it costs nothing.

Nothing happens on the cron URL

A 403 means the token is missing or wrong. A 500 means the run could not start, usually because no brand name is set or no questions exist. The endpoint returns JSON describing the run, so log the response body while you are setting it up.

Support

bilgi@aishortlist.app · Vera Bilişim